net/tipc/link.c

   1 /*
   2  * net/tipc/link.c: TIPC link code
   3  *
   4  * Copyright (c) 1996-2007, 2012-2014, Ericsson AB
   5  * Copyright (c) 2004-2007, 2010-2013, Wind River Systems
   6  * All rights reserved.
   7  *
   8  * Redistribution and use in source and binary forms, with or without
   9  * modification, are permitted provided that the following conditions are met:
  10  *
  11  * 1. Redistributions of source code must retain the above copyright
  12  *    notice, this list of conditions and the following disclaimer.
  13  * 2. Redistributions in binary form must reproduce the above copyright
  14  *    notice, this list of conditions and the following disclaimer in the
  15  *    documentation and/or other materials provided with the distribution.
  16  * 3. Neither the names of the copyright holders nor the names of its
  17  *    contributors may be used to endorse or promote products derived from
  18  *    this software without specific prior written permission.
  19  *
  20  * Alternatively, this software may be distributed under the terms of the
  21  * GNU General Public License ("GPL") version 2 as published by the Free
  22  * Software Foundation.
  23  *
  24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  25  * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  27  * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  28  * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  29  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  30  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  31  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  32  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  33  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  34  * POSSIBILITY OF SUCH DAMAGE.
  35  */
  36
  37 #include "core.h"
  38 #include "link.h"
  39 #include "port.h"
  40 #include "name_distr.h"
  41 #include "discover.h"
  42 #include "config.h"
  43
  44 #include <linux/pkt_sched.h>
  45
  46 /*
  47  * Error message prefixes
  48  */
  49 static const char *link_co_err = "Link changeover error, ";
  50 static const char *link_rst_msg = "Resetting link ";
  51 static const char *link_unk_evt = "Unknown link event ";
  52
  53 /*
  54  * Out-of-range value for link session numbers
  55  */
  56 #define INVALID_SESSION 0x10000
  57
  58 /*
  59  * Link state events:
  60  */
  61 #define  STARTING_EVT    856384768      /* link processing trigger */
  62 #define  TRAFFIC_MSG_EVT 560815u        /* rx'd ??? */
  63 #define  TIMEOUT_EVT     560817u        /* link timer expired */
  64
  65 /*
  66  * The following two 'message types' is really just implementation
  67  * data conveniently stored in the message header.
  68  * They must not be considered part of the protocol
  69  */
  70 #define OPEN_MSG   0
  71 #define CLOSED_MSG 1
  72
  73 /*
  74  * State value stored in 'exp_msg_count'
  75  */
  76 #define START_CHANGEOVER 100000u
  77
  78 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
  79                                        struct sk_buff *buf);
  80 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf);
  81 static int  tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
  82                                  struct sk_buff **buf);
  83 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance);
  84 static int  tipc_link_iovec_long_xmit(struct tipc_port *sender,
  85                                       struct iovec const *msg_sect,
  86                                       unsigned int len, u32 destnode);
  87 static void link_state_event(struct tipc_link *l_ptr, u32 event);
  88 static void link_reset_statistics(struct tipc_link *l_ptr);
  89 static void link_print(struct tipc_link *l_ptr, const char *str);
  90 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf);
  91 static void tipc_link_sync_xmit(struct tipc_link *l);
  92 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf);
  93
  94 /*
  95  *  Simple link routines
  96  */
  97 static unsigned int align(unsigned int i)
  98 {
  99         return (i + 3) & ~3u;
 100 }
 101
 102 static void link_init_max_pkt(struct tipc_link *l_ptr)
 103 {
 104         struct tipc_bearer *b_ptr;
 105         u32 max_pkt;
 106
 107         rcu_read_lock();
 108         b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
 109         if (!b_ptr) {
 110                 rcu_read_unlock();
 111                 return;
 112         }
 113         max_pkt = (b_ptr->mtu & ~3);
 114         rcu_read_unlock();
 115
 116         if (max_pkt > MAX_MSG_SIZE)
 117                 max_pkt = MAX_MSG_SIZE;
 118
 119         l_ptr->max_pkt_target = max_pkt;
 120         if (l_ptr->max_pkt_target < MAX_PKT_DEFAULT)
 121                 l_ptr->max_pkt = l_ptr->max_pkt_target;
 122         else
 123                 l_ptr->max_pkt = MAX_PKT_DEFAULT;
 124
 125         l_ptr->max_pkt_probes = 0;
 126 }
 127
 128 static u32 link_next_sent(struct tipc_link *l_ptr)
 129 {
 130         if (l_ptr->next_out)
 131                 return buf_seqno(l_ptr->next_out);
 132         return mod(l_ptr->next_out_no);
 133 }
 134
 135 static u32 link_last_sent(struct tipc_link *l_ptr)
 136 {
 137         return mod(link_next_sent(l_ptr) - 1);
 138 }
 139
 140 /*
 141  *  Simple non-static link routines (i.e. referenced outside this file)
 142  */
 143 int tipc_link_is_up(struct tipc_link *l_ptr)
 144 {
 145         if (!l_ptr)
 146                 return 0;
 147         return link_working_working(l_ptr) || link_working_unknown(l_ptr);
 148 }
 149
 150 int tipc_link_is_active(struct tipc_link *l_ptr)
 151 {
 152         return  (l_ptr->owner->active_links[0] == l_ptr) ||
 153                 (l_ptr->owner->active_links[1] == l_ptr);
 154 }
 155
 156 /**
 157  * link_timeout - handle expiration of link timer
 158  * @l_ptr: pointer to link
 159  */
 160 static void link_timeout(struct tipc_link *l_ptr)
 161 {
 162         tipc_node_lock(l_ptr->owner);
 163
 164         /* update counters used in statistical profiling of send traffic */
 165         l_ptr->stats.accu_queue_sz += l_ptr->out_queue_size;
 166         l_ptr->stats.queue_sz_counts++;
 167
 168         if (l_ptr->first_out) {
 169                 struct tipc_msg *msg = buf_msg(l_ptr->first_out);
 170                 u32 length = msg_size(msg);
 171
 172                 if ((msg_user(msg) == MSG_FRAGMENTER) &&
 173                     (msg_type(msg) == FIRST_FRAGMENT)) {
 174                         length = msg_size(msg_get_wrapped(msg));
 175                 }
 176                 if (length) {
 177                         l_ptr->stats.msg_lengths_total += length;
 178                         l_ptr->stats.msg_length_counts++;
 179                         if (length <= 64)
 180                                 l_ptr->stats.msg_length_profile[0]++;
 181                         else if (length <= 256)
 182                                 l_ptr->stats.msg_length_profile[1]++;
 183                         else if (length <= 1024)
 184                                 l_ptr->stats.msg_length_profile[2]++;
 185                         else if (length <= 4096)
 186                                 l_ptr->stats.msg_length_profile[3]++;
 187                         else if (length <= 16384)
 188                                 l_ptr->stats.msg_length_profile[4]++;
 189                         else if (length <= 32768)
 190                                 l_ptr->stats.msg_length_profile[5]++;
 191                         else
 192                                 l_ptr->stats.msg_length_profile[6]++;
 193                 }
 194         }
 195
 196         /* do all other link processing performed on a periodic basis */
 197
 198         link_state_event(l_ptr, TIMEOUT_EVT);
 199
 200         if (l_ptr->next_out)
 201                 tipc_link_push_queue(l_ptr);
 202
 203         tipc_node_unlock(l_ptr->owner);
 204 }
 205
 206 static void link_set_timer(struct tipc_link *l_ptr, u32 time)
 207 {
 208         k_start_timer(&l_ptr->timer, time);
 209 }
 210
 211 /**
 212  * tipc_link_create - create a new link
 213  * @n_ptr: pointer to associated node
 214  * @b_ptr: pointer to associated bearer
 215  * @media_addr: media address to use when sending messages over link
 216  *
 217  * Returns pointer to link.
 218  */
 219 struct tipc_link *tipc_link_create(struct tipc_node *n_ptr,
 220                                    struct tipc_bearer *b_ptr,
 221                                    const struct tipc_media_addr *media_addr)
 222 {
 223         struct tipc_link *l_ptr;
 224         struct tipc_msg *msg;
 225         char *if_name;
 226         char addr_string[16];
 227         u32 peer = n_ptr->addr;
 228
 229         if (n_ptr->link_cnt >= 2) {
 230                 tipc_addr_string_fill(addr_string, n_ptr->addr);
 231                 pr_err("Attempt to establish third link to %s\n", addr_string);
 232                 return NULL;
 233         }
 234
 235         if (n_ptr->links[b_ptr->identity]) {
 236                 tipc_addr_string_fill(addr_string, n_ptr->addr);
 237                 pr_err("Attempt to establish second link on <%s> to %s\n",
 238                        b_ptr->name, addr_string);
 239                 return NULL;
 240         }
 241
 242         l_ptr = kzalloc(sizeof(*l_ptr), GFP_ATOMIC);
 243         if (!l_ptr) {
 244                 pr_warn("Link creation failed, no memory\n");
 245                 return NULL;
 246         }
 247
 248         l_ptr->addr = peer;
 249         if_name = strchr(b_ptr->name, ':') + 1;
 250         sprintf(l_ptr->name, "%u.%u.%u:%s-%u.%u.%u:unknown",
 251                 tipc_zone(tipc_own_addr), tipc_cluster(tipc_own_addr),
 252                 tipc_node(tipc_own_addr),
 253                 if_name,
 254                 tipc_zone(peer), tipc_cluster(peer), tipc_node(peer));
 255                 /* note: peer i/f name is updated by reset/activate message */
 256         memcpy(&l_ptr->media_addr, media_addr, sizeof(*media_addr));
 257         l_ptr->owner = n_ptr;
 258         l_ptr->checkpoint = 1;
 259         l_ptr->peer_session = INVALID_SESSION;
 260         l_ptr->bearer_id = b_ptr->identity;
 261         link_set_supervision_props(l_ptr, b_ptr->tolerance);
 262         l_ptr->state = RESET_UNKNOWN;
 263
 264         l_ptr->pmsg = (struct tipc_msg *)&l_ptr->proto_msg;
 265         msg = l_ptr->pmsg;
 266         tipc_msg_init(msg, LINK_PROTOCOL, RESET_MSG, INT_H_SIZE, l_ptr->addr);
 267         msg_set_size(msg, sizeof(l_ptr->proto_msg));
 268         msg_set_session(msg, (tipc_random & 0xffff));
 269         msg_set_bearer_id(msg, b_ptr->identity);
 270         strcpy((char *)msg_data(msg), if_name);
 271
 272         l_ptr->priority = b_ptr->priority;
 273         tipc_link_set_queue_limits(l_ptr, b_ptr->window);
 274
 275         l_ptr->net_plane = b_ptr->net_plane;
 276         link_init_max_pkt(l_ptr);
 277
 278         l_ptr->next_out_no = 1;
 279         INIT_LIST_HEAD(&l_ptr->waiting_ports);
 280
 281         link_reset_statistics(l_ptr);
 282
 283         tipc_node_attach_link(n_ptr, l_ptr);
 284
 285         k_init_timer(&l_ptr->timer, (Handler)link_timeout,
 286                      (unsigned long)l_ptr);
 287
 288         link_state_event(l_ptr, STARTING_EVT);
 289
 290         return l_ptr;
 291 }
 292
 293 void tipc_link_delete_list(unsigned int bearer_id, bool shutting_down)
 294 {
 295         struct tipc_link *l_ptr;
 296         struct tipc_node *n_ptr;
 297
 298         rcu_read_lock();
 299         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
 300                 tipc_node_lock(n_ptr);
 301                 l_ptr = n_ptr->links[bearer_id];
 302                 if (l_ptr) {
 303                         tipc_link_reset(l_ptr);
 304                         if (shutting_down || !tipc_node_is_up(n_ptr)) {
 305                                 tipc_node_detach_link(l_ptr->owner, l_ptr);
 306                                 tipc_link_reset_fragments(l_ptr);
 307                                 tipc_node_unlock(n_ptr);
 308
 309                                 /* Nobody else can access this link now: */
 310                                 del_timer_sync(&l_ptr->timer);
 311                                 kfree(l_ptr);
 312                         } else {
 313                                 /* Detach/delete when failover is finished: */
 314                                 l_ptr->flags |= LINK_STOPPED;
 315                                 tipc_node_unlock(n_ptr);
 316                                 del_timer_sync(&l_ptr->timer);
 317                         }
 318                         continue;
 319                 }
 320                 tipc_node_unlock(n_ptr);
 321         }
 322         rcu_read_unlock();
 323 }
 324
 325 /**
 326  * link_schedule_port - schedule port for deferred sending
 327  * @l_ptr: pointer to link
 328  * @origport: reference to sending port
 329  * @sz: amount of data to be sent
 330  *
 331  * Schedules port for renewed sending of messages after link congestion
 332  * has abated.
 333  */
 334 static int link_schedule_port(struct tipc_link *l_ptr, u32 origport, u32 sz)
 335 {
 336         struct tipc_port *p_ptr;
 337
 338         spin_lock_bh(&tipc_port_list_lock);
 339         p_ptr = tipc_port_lock(origport);
 340         if (p_ptr) {
 341                 if (!list_empty(&p_ptr->wait_list))
 342                         goto exit;
 343                 p_ptr->congested = 1;
 344                 p_ptr->waiting_pkts = 1 + ((sz - 1) / l_ptr->max_pkt);
 345                 list_add_tail(&p_ptr->wait_list, &l_ptr->waiting_ports);
 346                 l_ptr->stats.link_congs++;
 347 exit:
 348                 tipc_port_unlock(p_ptr);
 349         }
 350         spin_unlock_bh(&tipc_port_list_lock);
 351         return -ELINKCONG;
 352 }
 353
 354 void tipc_link_wakeup_ports(struct tipc_link *l_ptr, int all)
 355 {
 356         struct tipc_port *p_ptr;
 357         struct tipc_port *temp_p_ptr;
 358         int win = l_ptr->queue_limit[0] - l_ptr->out_queue_size;
 359
 360         if (all)
 361                 win = 100000;
 362         if (win <= 0)
 363                 return;
 364         if (!spin_trylock_bh(&tipc_port_list_lock))
 365                 return;
 366         if (link_congested(l_ptr))
 367                 goto exit;
 368         list_for_each_entry_safe(p_ptr, temp_p_ptr, &l_ptr->waiting_ports,
 369                                  wait_list) {
 370                 if (win <= 0)
 371                         break;
 372                 list_del_init(&p_ptr->wait_list);
 373                 spin_lock_bh(p_ptr->lock);
 374                 p_ptr->congested = 0;
 375                 tipc_port_wakeup(p_ptr);
 376                 win -= p_ptr->waiting_pkts;
 377                 spin_unlock_bh(p_ptr->lock);
 378         }
 379
 380 exit:
 381         spin_unlock_bh(&tipc_port_list_lock);
 382 }
 383
 384 /**
 385  * link_release_outqueue - purge link's outbound message queue
 386  * @l_ptr: pointer to link
 387  */
 388 static void link_release_outqueue(struct tipc_link *l_ptr)
 389 {
 390         kfree_skb_list(l_ptr->first_out);
 391         l_ptr->first_out = NULL;
 392         l_ptr->out_queue_size = 0;
 393 }
 394
 395 /**
 396  * tipc_link_reset_fragments - purge link's inbound message fragments queue
 397  * @l_ptr: pointer to link
 398  */
 399 void tipc_link_reset_fragments(struct tipc_link *l_ptr)
 400 {
 401         kfree_skb(l_ptr->reasm_head);
 402         l_ptr->reasm_head = NULL;
 403         l_ptr->reasm_tail = NULL;
 404 }
 405
 406 /**
 407  * tipc_link_purge_queues - purge all pkt queues associated with link
 408  * @l_ptr: pointer to link
 409  */
 410 void tipc_link_purge_queues(struct tipc_link *l_ptr)
 411 {
 412         kfree_skb_list(l_ptr->oldest_deferred_in);
 413         kfree_skb_list(l_ptr->first_out);
 414         tipc_link_reset_fragments(l_ptr);
 415         kfree_skb(l_ptr->proto_msg_queue);
 416         l_ptr->proto_msg_queue = NULL;
 417 }
 418
 419 void tipc_link_reset(struct tipc_link *l_ptr)
 420 {
 421         u32 prev_state = l_ptr->state;
 422         u32 checkpoint = l_ptr->next_in_no;
 423         int was_active_link = tipc_link_is_active(l_ptr);
 424
 425         msg_set_session(l_ptr->pmsg, ((msg_session(l_ptr->pmsg) + 1) & 0xffff));
 426
 427         /* Link is down, accept any session */
 428         l_ptr->peer_session = INVALID_SESSION;
 429
 430         /* Prepare for max packet size negotiation */
 431         link_init_max_pkt(l_ptr);
 432
 433         l_ptr->state = RESET_UNKNOWN;
 434
 435         if ((prev_state == RESET_UNKNOWN) || (prev_state == RESET_RESET))
 436                 return;
 437
 438         tipc_node_link_down(l_ptr->owner, l_ptr);
 439         tipc_bearer_remove_dest(l_ptr->bearer_id, l_ptr->addr);
 440
 441         if (was_active_link && tipc_node_active_links(l_ptr->owner)) {
 442                 l_ptr->reset_checkpoint = checkpoint;
 443                 l_ptr->exp_msg_count = START_CHANGEOVER;
 444         }
 445
 446         /* Clean up all queues: */
 447         link_release_outqueue(l_ptr);
 448         kfree_skb(l_ptr->proto_msg_queue);
 449         l_ptr->proto_msg_queue = NULL;
 450         kfree_skb_list(l_ptr->oldest_deferred_in);
 451         if (!list_empty(&l_ptr->waiting_ports))
 452                 tipc_link_wakeup_ports(l_ptr, 1);
 453
 454         l_ptr->retransm_queue_head = 0;
 455         l_ptr->retransm_queue_size = 0;
 456         l_ptr->last_out = NULL;
 457         l_ptr->first_out = NULL;
 458         l_ptr->next_out = NULL;
 459         l_ptr->unacked_window = 0;
 460         l_ptr->checkpoint = 1;
 461         l_ptr->next_out_no = 1;
 462         l_ptr->deferred_inqueue_sz = 0;
 463         l_ptr->oldest_deferred_in = NULL;
 464         l_ptr->newest_deferred_in = NULL;
 465         l_ptr->fsm_msg_cnt = 0;
 466         l_ptr->stale_count = 0;
 467         link_reset_statistics(l_ptr);
 468 }
 469
 470 void tipc_link_reset_list(unsigned int bearer_id)
 471 {
 472         struct tipc_link *l_ptr;
 473         struct tipc_node *n_ptr;
 474
 475         rcu_read_lock();
 476         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
 477                 tipc_node_lock(n_ptr);
 478                 l_ptr = n_ptr->links[bearer_id];
 479                 if (l_ptr)
 480                         tipc_link_reset(l_ptr);
 481                 tipc_node_unlock(n_ptr);
 482         }
 483         rcu_read_unlock();
 484 }
 485
 486 static void link_activate(struct tipc_link *l_ptr)
 487 {
 488         l_ptr->next_in_no = l_ptr->stats.recv_info = 1;
 489         tipc_node_link_up(l_ptr->owner, l_ptr);
 490         tipc_bearer_add_dest(l_ptr->bearer_id, l_ptr->addr);
 491 }
 492
 493 /**
 494  * link_state_event - link finite state machine
 495  * @l_ptr: pointer to link
 496  * @event: state machine event to process
 497  */
 498 static void link_state_event(struct tipc_link *l_ptr, unsigned int event)
 499 {
 500         struct tipc_link *other;
 501         u32 cont_intv = l_ptr->continuity_interval;
 502
 503         if (l_ptr->flags & LINK_STOPPED)
 504                 return;
 505
 506         if (!(l_ptr->flags & LINK_STARTED) && (event != STARTING_EVT))
 507                 return;         /* Not yet. */
 508
 509         /* Check whether changeover is going on */
 510         if (l_ptr->exp_msg_count) {
 511                 if (event == TIMEOUT_EVT)
 512                         link_set_timer(l_ptr, cont_intv);
 513                 return;
 514         }
 515
 516         switch (l_ptr->state) {
 517         case WORKING_WORKING:
 518                 switch (event) {
 519                 case TRAFFIC_MSG_EVT:
 520                 case ACTIVATE_MSG:
 521                         break;
 522                 case TIMEOUT_EVT:
 523                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 524                                 l_ptr->checkpoint = l_ptr->next_in_no;
 525                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 526                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 527                                                              0, 0, 0, 0, 0);
 528                                         l_ptr->fsm_msg_cnt++;
 529                                 } else if (l_ptr->max_pkt < l_ptr->max_pkt_target) {
 530                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 531                                                              1, 0, 0, 0, 0);
 532                                         l_ptr->fsm_msg_cnt++;
 533                                 }
 534                                 link_set_timer(l_ptr, cont_intv);
 535                                 break;
 536                         }
 537                         l_ptr->state = WORKING_UNKNOWN;
 538                         l_ptr->fsm_msg_cnt = 0;
 539                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 540                         l_ptr->fsm_msg_cnt++;
 541                         link_set_timer(l_ptr, cont_intv / 4);
 542                         break;
 543                 case RESET_MSG:
 544                         pr_info("%s<%s>, requested by peer\n", link_rst_msg,
 545                                 l_ptr->name);
 546                         tipc_link_reset(l_ptr);
 547                         l_ptr->state = RESET_RESET;
 548                         l_ptr->fsm_msg_cnt = 0;
 549                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 550                                              0, 0, 0, 0, 0);
 551                         l_ptr->fsm_msg_cnt++;
 552                         link_set_timer(l_ptr, cont_intv);
 553                         break;
 554                 default:
 555                         pr_err("%s%u in WW state\n", link_unk_evt, event);
 556                 }
 557                 break;
 558         case WORKING_UNKNOWN:
 559                 switch (event) {
 560                 case TRAFFIC_MSG_EVT:
 561                 case ACTIVATE_MSG:
 562                         l_ptr->state = WORKING_WORKING;
 563                         l_ptr->fsm_msg_cnt = 0;
 564                         link_set_timer(l_ptr, cont_intv);
 565                         break;
 566                 case RESET_MSG:
 567                         pr_info("%s<%s>, requested by peer while probing\n",
 568                                 link_rst_msg, l_ptr->name);
 569                         tipc_link_reset(l_ptr);
 570                         l_ptr->state = RESET_RESET;
 571                         l_ptr->fsm_msg_cnt = 0;
 572                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 573                                              0, 0, 0, 0, 0);
 574                         l_ptr->fsm_msg_cnt++;
 575                         link_set_timer(l_ptr, cont_intv);
 576                         break;
 577                 case TIMEOUT_EVT:
 578                         if (l_ptr->next_in_no != l_ptr->checkpoint) {
 579                                 l_ptr->state = WORKING_WORKING;
 580                                 l_ptr->fsm_msg_cnt = 0;
 581                                 l_ptr->checkpoint = l_ptr->next_in_no;
 582                                 if (tipc_bclink_acks_missing(l_ptr->owner)) {
 583                                         tipc_link_proto_xmit(l_ptr, STATE_MSG,
 584                                                              0, 0, 0, 0, 0);
 585                                         l_ptr->fsm_msg_cnt++;
 586                                 }
 587                                 link_set_timer(l_ptr, cont_intv);
 588                         } else if (l_ptr->fsm_msg_cnt < l_ptr->abort_limit) {
 589                                 tipc_link_proto_xmit(l_ptr, STATE_MSG,
 590                                                      1, 0, 0, 0, 0);
 591                                 l_ptr->fsm_msg_cnt++;
 592                                 link_set_timer(l_ptr, cont_intv / 4);
 593                         } else {        /* Link has failed */
 594                                 pr_warn("%s<%s>, peer not responding\n",
 595                                         link_rst_msg, l_ptr->name);
 596                                 tipc_link_reset(l_ptr);
 597                                 l_ptr->state = RESET_UNKNOWN;
 598                                 l_ptr->fsm_msg_cnt = 0;
 599                                 tipc_link_proto_xmit(l_ptr, RESET_MSG,
 600                                                      0, 0, 0, 0, 0);
 601                                 l_ptr->fsm_msg_cnt++;
 602                                 link_set_timer(l_ptr, cont_intv);
 603                         }
 604                         break;
 605                 default:
 606                         pr_err("%s%u in WU state\n", link_unk_evt, event);
 607                 }
 608                 break;
 609         case RESET_UNKNOWN:
 610                 switch (event) {
 611                 case TRAFFIC_MSG_EVT:
 612                         break;
 613                 case ACTIVATE_MSG:
 614                         other = l_ptr->owner->active_links[0];
 615                         if (other && link_working_unknown(other))
 616                                 break;
 617                         l_ptr->state = WORKING_WORKING;
 618                         l_ptr->fsm_msg_cnt = 0;
 619                         link_activate(l_ptr);
 620                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 621                         l_ptr->fsm_msg_cnt++;
 622                         if (l_ptr->owner->working_links == 1)
 623                                 tipc_link_sync_xmit(l_ptr);
 624                         link_set_timer(l_ptr, cont_intv);
 625                         break;
 626                 case RESET_MSG:
 627                         l_ptr->state = RESET_RESET;
 628                         l_ptr->fsm_msg_cnt = 0;
 629                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 630                                              1, 0, 0, 0, 0);
 631                         l_ptr->fsm_msg_cnt++;
 632                         link_set_timer(l_ptr, cont_intv);
 633                         break;
 634                 case STARTING_EVT:
 635                         l_ptr->flags |= LINK_STARTED;
 636                         /* fall through */
 637                 case TIMEOUT_EVT:
 638                         tipc_link_proto_xmit(l_ptr, RESET_MSG, 0, 0, 0, 0, 0);
 639                         l_ptr->fsm_msg_cnt++;
 640                         link_set_timer(l_ptr, cont_intv);
 641                         break;
 642                 default:
 643                         pr_err("%s%u in RU state\n", link_unk_evt, event);
 644                 }
 645                 break;
 646         case RESET_RESET:
 647                 switch (event) {
 648                 case TRAFFIC_MSG_EVT:
 649                 case ACTIVATE_MSG:
 650                         other = l_ptr->owner->active_links[0];
 651                         if (other && link_working_unknown(other))
 652                                 break;
 653                         l_ptr->state = WORKING_WORKING;
 654                         l_ptr->fsm_msg_cnt = 0;
 655                         link_activate(l_ptr);
 656                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 1, 0, 0, 0, 0);
 657                         l_ptr->fsm_msg_cnt++;
 658                         if (l_ptr->owner->working_links == 1)
 659                                 tipc_link_sync_xmit(l_ptr);
 660                         link_set_timer(l_ptr, cont_intv);
 661                         break;
 662                 case RESET_MSG:
 663                         break;
 664                 case TIMEOUT_EVT:
 665                         tipc_link_proto_xmit(l_ptr, ACTIVATE_MSG,
 666                                              0, 0, 0, 0, 0);
 667                         l_ptr->fsm_msg_cnt++;
 668                         link_set_timer(l_ptr, cont_intv);
 669                         break;
 670                 default:
 671                         pr_err("%s%u in RR state\n", link_unk_evt, event);
 672                 }
 673                 break;
 674         default:
 675                 pr_err("Unknown link state %u/%u\n", l_ptr->state, event);
 676         }
 677 }
 678
 679 /*
 680  * link_bundle_buf(): Append contents of a buffer to
 681  * the tail of an existing one.
 682  */
 683 static int link_bundle_buf(struct tipc_link *l_ptr, struct sk_buff *bundler,
 684                            struct sk_buff *buf)
 685 {
 686         struct tipc_msg *bundler_msg = buf_msg(bundler);
 687         struct tipc_msg *msg = buf_msg(buf);
 688         u32 size = msg_size(msg);
 689         u32 bundle_size = msg_size(bundler_msg);
 690         u32 to_pos = align(bundle_size);
 691         u32 pad = to_pos - bundle_size;
 692
 693         if (msg_user(bundler_msg) != MSG_BUNDLER)
 694                 return 0;
 695         if (msg_type(bundler_msg) != OPEN_MSG)
 696                 return 0;
 697         if (skb_tailroom(bundler) < (pad + size))
 698                 return 0;
 699         if (l_ptr->max_pkt < (to_pos + size))
 700                 return 0;
 701
 702         skb_put(bundler, pad + size);
 703         skb_copy_to_linear_data_offset(bundler, to_pos, buf->data, size);
 704         msg_set_size(bundler_msg, to_pos + size);
 705         msg_set_msgcnt(bundler_msg, msg_msgcnt(bundler_msg) + 1);
 706         kfree_skb(buf);
 707         l_ptr->stats.sent_bundled++;
 708         return 1;
 709 }
 710
 711 static void link_add_to_outqueue(struct tipc_link *l_ptr,
 712                                  struct sk_buff *buf,
 713                                  struct tipc_msg *msg)
 714 {
 715         u32 ack = mod(l_ptr->next_in_no - 1);
 716         u32 seqno = mod(l_ptr->next_out_no++);
 717
 718         msg_set_word(msg, 2, ((ack << 16) | seqno));
 719         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
 720         buf->next = NULL;
 721         if (l_ptr->first_out) {
 722                 l_ptr->last_out->next = buf;
 723                 l_ptr->last_out = buf;
 724         } else
 725                 l_ptr->first_out = l_ptr->last_out = buf;
 726
 727         l_ptr->out_queue_size++;
 728         if (l_ptr->out_queue_size > l_ptr->stats.max_queue_sz)
 729                 l_ptr->stats.max_queue_sz = l_ptr->out_queue_size;
 730 }
 731
 732 static void link_add_chain_to_outqueue(struct tipc_link *l_ptr,
 733                                        struct sk_buff *buf_chain,
 734                                        u32 long_msgno)
 735 {
 736         struct sk_buff *buf;
 737         struct tipc_msg *msg;
 738
 739         if (!l_ptr->next_out)
 740                 l_ptr->next_out = buf_chain;
 741         while (buf_chain) {
 742                 buf = buf_chain;
 743                 buf_chain = buf_chain->next;
 744
 745                 msg = buf_msg(buf);
 746                 msg_set_long_msgno(msg, long_msgno);
 747                 link_add_to_outqueue(l_ptr, buf, msg);
 748         }
 749 }
 750
 751 /*
 752  * tipc_link_xmit() is the 'full path' for messages, called from
 753  * inside TIPC when the 'fast path' in tipc_send_xmit
 754  * has failed, and from link_send()
 755  */
 756 int __tipc_link_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
 757 {
 758         struct tipc_msg *msg = buf_msg(buf);
 759         u32 size = msg_size(msg);
 760         u32 dsz = msg_data_sz(msg);
 761         u32 queue_size = l_ptr->out_queue_size;
 762         u32 imp = tipc_msg_tot_importance(msg);
 763         u32 queue_limit = l_ptr->queue_limit[imp];
 764         u32 max_packet = l_ptr->max_pkt;
 765
 766         /* Match msg importance against queue limits: */
 767         if (unlikely(queue_size >= queue_limit)) {
 768                 if (imp <= TIPC_CRITICAL_IMPORTANCE) {
 769                         link_schedule_port(l_ptr, msg_origport(msg), size);
 770                         kfree_skb(buf);
 771                         return -ELINKCONG;
 772                 }
 773                 kfree_skb(buf);
 774                 if (imp > CONN_MANAGER) {
 775                         pr_warn("%s<%s>, send queue full", link_rst_msg,
 776                                 l_ptr->name);
 777                         tipc_link_reset(l_ptr);
 778                 }
 779                 return dsz;
 780         }
 781
 782         /* Fragmentation needed ? */
 783         if (size > max_packet)
 784                 return tipc_link_frag_xmit(l_ptr, buf);
 785
 786         /* Packet can be queued or sent. */
 787         if (likely(!link_congested(l_ptr))) {
 788                 link_add_to_outqueue(l_ptr, buf, msg);
 789
 790                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
 791                 l_ptr->unacked_window = 0;
 792                 return dsz;
 793         }
 794         /* Congestion: can message be bundled ? */
 795         if ((msg_user(msg) != CHANGEOVER_PROTOCOL) &&
 796             (msg_user(msg) != MSG_FRAGMENTER)) {
 797
 798                 /* Try adding message to an existing bundle */
 799                 if (l_ptr->next_out &&
 800                     link_bundle_buf(l_ptr, l_ptr->last_out, buf))
 801                         return dsz;
 802
 803                 /* Try creating a new bundle */
 804                 if (size <= max_packet * 2 / 3) {
 805                         struct sk_buff *bundler = tipc_buf_acquire(max_packet);
 806                         struct tipc_msg bundler_hdr;
 807
 808                         if (bundler) {
 809                                 tipc_msg_init(&bundler_hdr, MSG_BUNDLER, OPEN_MSG,
 810                                          INT_H_SIZE, l_ptr->addr);
 811                                 skb_copy_to_linear_data(bundler, &bundler_hdr,
 812                                                         INT_H_SIZE);
 813                                 skb_trim(bundler, INT_H_SIZE);
 814                                 link_bundle_buf(l_ptr, bundler, buf);
 815                                 buf = bundler;
 816                                 msg = buf_msg(buf);
 817                                 l_ptr->stats.sent_bundles++;
 818                         }
 819                 }
 820         }
 821         if (!l_ptr->next_out)
 822                 l_ptr->next_out = buf;
 823         link_add_to_outqueue(l_ptr, buf, msg);
 824         return dsz;
 825 }
 826
 827 /*
 828  * tipc_link_xmit(): same as __tipc_link_xmit(), but the link to use
 829  * has not been selected yet, and the the owner node is not locked
 830  * Called by TIPC internal users, e.g. the name distributor
 831  */
 832 int tipc_link_xmit(struct sk_buff *buf, u32 dest, u32 selector)
 833 {
 834         struct tipc_link *l_ptr;
 835         struct tipc_node *n_ptr;
 836         int res = -ELINKCONG;
 837
 838         n_ptr = tipc_node_find(dest);
 839         if (n_ptr) {
 840                 tipc_node_lock(n_ptr);
 841                 l_ptr = n_ptr->active_links[selector & 1];
 842                 if (l_ptr)
 843                         res = __tipc_link_xmit(l_ptr, buf);
 844                 else
 845                         kfree_skb(buf);
 846                 tipc_node_unlock(n_ptr);
 847         } else {
 848                 kfree_skb(buf);
 849         }
 850         return res;
 851 }
 852
 853 /*
 854  * tipc_link_sync_xmit - synchronize broadcast link endpoints.
 855  *
 856  * Give a newly added peer node the sequence number where it should
 857  * start receiving and acking broadcast packets.
 858  *
 859  * Called with node locked
 860  */
 861 static void tipc_link_sync_xmit(struct tipc_link *l)
 862 {
 863         struct sk_buff *buf;
 864         struct tipc_msg *msg;
 865
 866         buf = tipc_buf_acquire(INT_H_SIZE);
 867         if (!buf)
 868                 return;
 869
 870         msg = buf_msg(buf);
 871         tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG, INT_H_SIZE, l->addr);
 872         msg_set_last_bcast(msg, l->owner->bclink.acked);
 873         link_add_chain_to_outqueue(l, buf, 0);
 874         tipc_link_push_queue(l);
 875 }
 876
 877 /*
 878  * tipc_link_sync_rcv - synchronize broadcast link endpoints.
 879  * Receive the sequence number where we should start receiving and
 880  * acking broadcast packets from a newly added peer node, and open
 881  * up for reception of such packets.
 882  *
 883  * Called with node locked
 884  */
 885 static void tipc_link_sync_rcv(struct tipc_node *n, struct sk_buff *buf)
 886 {
 887         struct tipc_msg *msg = buf_msg(buf);
 888
 889         n->bclink.last_sent = n->bclink.last_in = msg_last_bcast(msg);
 890         n->bclink.recv_permitted = true;
 891         kfree_skb(buf);
 892 }
 893
 894 /*
 895  * tipc_link_names_xmit - send name table entries to new neighbor
 896  *
 897  * Send routine for bulk delivery of name table messages when contact
 898  * with a new neighbor occurs. No link congestion checking is performed
 899  * because name table messages *must* be delivered. The messages must be
 900  * small enough not to require fragmentation.
 901  * Called without any locks held.
 902  */
 903 void tipc_link_names_xmit(struct list_head *message_list, u32 dest)
 904 {
 905         struct tipc_node *n_ptr;
 906         struct tipc_link *l_ptr;
 907         struct sk_buff *buf;
 908         struct sk_buff *temp_buf;
 909
 910         if (list_empty(message_list))
 911                 return;
 912
 913         n_ptr = tipc_node_find(dest);
 914         if (n_ptr) {
 915                 tipc_node_lock(n_ptr);
 916                 l_ptr = n_ptr->active_links[0];
 917                 if (l_ptr) {
 918                         /* convert circular list to linear list */
 919                         ((struct sk_buff *)message_list->prev)->next = NULL;
 920                         link_add_chain_to_outqueue(l_ptr,
 921                                 (struct sk_buff *)message_list->next, 0);
 922                         tipc_link_push_queue(l_ptr);
 923                         INIT_LIST_HEAD(message_list);
 924                 }
 925                 tipc_node_unlock(n_ptr);
 926         }
 927
 928         /* discard the messages if they couldn't be sent */
 929         list_for_each_safe(buf, temp_buf, ((struct sk_buff *)message_list)) {
 930                 list_del((struct list_head *)buf);
 931                 kfree_skb(buf);
 932         }
 933 }
 934
 935 /*
 936  * tipc_link_xmit_fast: Entry for data messages where the
 937  * destination link is known and the header is complete,
 938  * inclusive total message length. Very time critical.
 939  * Link is locked. Returns user data length.
 940  */
 941 static int tipc_link_xmit_fast(struct tipc_link *l_ptr, struct sk_buff *buf,
 942                                u32 *used_max_pkt)
 943 {
 944         struct tipc_msg *msg = buf_msg(buf);
 945         int res = msg_data_sz(msg);
 946
 947         if (likely(!link_congested(l_ptr))) {
 948                 if (likely(msg_size(msg) <= l_ptr->max_pkt)) {
 949                         link_add_to_outqueue(l_ptr, buf, msg);
 950                         tipc_bearer_send(l_ptr->bearer_id, buf,
 951                                          &l_ptr->media_addr);
 952                         l_ptr->unacked_window = 0;
 953                         return res;
 954                 }
 955                 else
 956                         *used_max_pkt = l_ptr->max_pkt;
 957         }
 958         return __tipc_link_xmit(l_ptr, buf);  /* All other cases */
 959 }
 960
 961 /*
 962  * tipc_link_iovec_xmit_fast: Entry for messages where the
 963  * destination processor is known and the header is complete,
 964  * except for total message length.
 965  * Returns user data length or errno.
 966  */
 967 int tipc_link_iovec_xmit_fast(struct tipc_port *sender,
 968                               struct iovec const *msg_sect,
 969                               unsigned int len, u32 destaddr)
 970 {
 971         struct tipc_msg *hdr = &sender->phdr;
 972         struct tipc_link *l_ptr;
 973         struct sk_buff *buf;
 974         struct tipc_node *node;
 975         int res;
 976         u32 selector = msg_origport(hdr) & 1;
 977
 978 again:
 979         /*
 980          * Try building message using port's max_pkt hint.
 981          * (Must not hold any locks while building message.)
 982          */
 983         res = tipc_msg_build(hdr, msg_sect, len, sender->max_pkt, &buf);
 984         /* Exit if build request was invalid */
 985         if (unlikely(res < 0))
 986                 return res;
 987
 988         node = tipc_node_find(destaddr);
 989         if (likely(node)) {
 990                 tipc_node_lock(node);
 991                 l_ptr = node->active_links[selector];
 992                 if (likely(l_ptr)) {
 993                         if (likely(buf)) {
 994                                 res = tipc_link_xmit_fast(l_ptr, buf,
 995                                                           &sender->max_pkt);
 996 exit:
 997                                 tipc_node_unlock(node);
 998                                 return res;
 999                         }
1000
1001                         /* Exit if link (or bearer) is congested */
1002                         if (link_congested(l_ptr)) {
1003                                 res = link_schedule_port(l_ptr,
1004                                                          sender->ref, res);
1005                                 goto exit;
1006                         }
1007
1008                         /*
1009                          * Message size exceeds max_pkt hint; update hint,
1010                          * then re-try fast path or fragment the message
1011                          */
1012                         sender->max_pkt = l_ptr->max_pkt;
1013                         tipc_node_unlock(node);
1014
1015
1016                         if ((msg_hdr_sz(hdr) + res) <= sender->max_pkt)
1017                                 goto again;
1018
1019                         return tipc_link_iovec_long_xmit(sender, msg_sect,
1020                                                          len, destaddr);
1021                 }
1022                 tipc_node_unlock(node);
1023         }
1024
1025         /* Couldn't find a link to the destination node */
1026         kfree_skb(buf);
1027         tipc_port_iovec_reject(sender, hdr, msg_sect, len, TIPC_ERR_NO_NODE);
1028         return -ENETUNREACH;
1029 }
1030
1031 /*
1032  * tipc_link_iovec_long_xmit(): Entry for long messages where the
1033  * destination node is known and the header is complete,
1034  * inclusive total message length.
1035  * Link and bearer congestion status have been checked to be ok,
1036  * and are ignored if they change.
1037  *
1038  * Note that fragments do not use the full link MTU so that they won't have
1039  * to undergo refragmentation if link changeover causes them to be sent
1040  * over another link with an additional tunnel header added as prefix.
1041  * (Refragmentation will still occur if the other link has a smaller MTU.)
1042  *
1043  * Returns user data length or errno.
1044  */
1045 static int tipc_link_iovec_long_xmit(struct tipc_port *sender,
1046                                      struct iovec const *msg_sect,
1047                                      unsigned int len, u32 destaddr)
1048 {
1049         struct tipc_link *l_ptr;
1050         struct tipc_node *node;
1051         struct tipc_msg *hdr = &sender->phdr;
1052         u32 dsz = len;
1053         u32 max_pkt, fragm_sz, rest;
1054         struct tipc_msg fragm_hdr;
1055         struct sk_buff *buf, *buf_chain, *prev;
1056         u32 fragm_crs, fragm_rest, hsz, sect_rest;
1057         const unchar __user *sect_crs;
1058         int curr_sect;
1059         u32 fragm_no;
1060         int res = 0;
1061
1062 again:
1063         fragm_no = 1;
1064         max_pkt = sender->max_pkt - INT_H_SIZE;
1065                 /* leave room for tunnel header in case of link changeover */
1066         fragm_sz = max_pkt - INT_H_SIZE;
1067                 /* leave room for fragmentation header in each fragment */
1068         rest = dsz;
1069         fragm_crs = 0;
1070         fragm_rest = 0;
1071         sect_rest = 0;
1072         sect_crs = NULL;
1073         curr_sect = -1;
1074
1075         /* Prepare reusable fragment header */
1076         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
1077                  INT_H_SIZE, msg_destnode(hdr));
1078         msg_set_size(&fragm_hdr, max_pkt);
1079         msg_set_fragm_no(&fragm_hdr, 1);
1080
1081         /* Prepare header of first fragment */
1082         buf_chain = buf = tipc_buf_acquire(max_pkt);
1083         if (!buf)
1084                 return -ENOMEM;
1085         buf->next = NULL;
1086         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1087         hsz = msg_hdr_sz(hdr);
1088         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, hdr, hsz);
1089
1090         /* Chop up message */
1091         fragm_crs = INT_H_SIZE + hsz;
1092         fragm_rest = fragm_sz - hsz;
1093
1094         do {            /* For all sections */
1095                 u32 sz;
1096
1097                 if (!sect_rest) {
1098                         sect_rest = msg_sect[++curr_sect].iov_len;
1099                         sect_crs = msg_sect[curr_sect].iov_base;
1100                 }
1101
1102                 if (sect_rest < fragm_rest)
1103                         sz = sect_rest;
1104                 else
1105                         sz = fragm_rest;
1106
1107                 if (copy_from_user(buf->data + fragm_crs, sect_crs, sz)) {
1108                         res = -EFAULT;
1109 error:
1110                         kfree_skb_list(buf_chain);
1111                         return res;
1112                 }
1113                 sect_crs += sz;
1114                 sect_rest -= sz;
1115                 fragm_crs += sz;
1116                 fragm_rest -= sz;
1117                 rest -= sz;
1118
1119                 if (!fragm_rest && rest) {
1120
1121                         /* Initiate new fragment: */
1122                         if (rest <= fragm_sz) {
1123                                 fragm_sz = rest;
1124                                 msg_set_type(&fragm_hdr, LAST_FRAGMENT);
1125                         } else {
1126                                 msg_set_type(&fragm_hdr, FRAGMENT);
1127                         }
1128                         msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
1129                         msg_set_fragm_no(&fragm_hdr, ++fragm_no);
1130                         prev = buf;
1131                         buf = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
1132                         if (!buf) {
1133                                 res = -ENOMEM;
1134                                 goto error;
1135                         }
1136
1137                         buf->next = NULL;
1138                         prev->next = buf;
1139                         skb_copy_to_linear_data(buf, &fragm_hdr, INT_H_SIZE);
1140                         fragm_crs = INT_H_SIZE;
1141                         fragm_rest = fragm_sz;
1142                 }
1143         } while (rest > 0);
1144
1145         /*
1146          * Now we have a buffer chain. Select a link and check
1147          * that packet size is still OK
1148          */
1149         node = tipc_node_find(destaddr);
1150         if (likely(node)) {
1151                 tipc_node_lock(node);
1152                 l_ptr = node->active_links[sender->ref & 1];
1153                 if (!l_ptr) {
1154                         tipc_node_unlock(node);
1155                         goto reject;
1156                 }
1157                 if (l_ptr->max_pkt < max_pkt) {
1158                         sender->max_pkt = l_ptr->max_pkt;
1159                         tipc_node_unlock(node);
1160                         kfree_skb_list(buf_chain);
1161                         goto again;
1162                 }
1163         } else {
1164 reject:
1165                 kfree_skb_list(buf_chain);
1166                 tipc_port_iovec_reject(sender, hdr, msg_sect, len,
1167                                        TIPC_ERR_NO_NODE);
1168                 return -ENETUNREACH;
1169         }
1170
1171         /* Append chain of fragments to send queue & send them */
1172         l_ptr->long_msg_seq_no++;
1173         link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
1174         l_ptr->stats.sent_fragments += fragm_no;
1175         l_ptr->stats.sent_fragmented++;
1176         tipc_link_push_queue(l_ptr);
1177         tipc_node_unlock(node);
1178         return dsz;
1179 }
1180
1181 /*
1182  * tipc_link_push_packet: Push one unsent packet to the media
1183  */
1184 static u32 tipc_link_push_packet(struct tipc_link *l_ptr)
1185 {
1186         struct sk_buff *buf = l_ptr->first_out;
1187         u32 r_q_size = l_ptr->retransm_queue_size;
1188         u32 r_q_head = l_ptr->retransm_queue_head;
1189
1190         /* Step to position where retransmission failed, if any,    */
1191         /* consider that buffers may have been released in meantime */
1192         if (r_q_size && buf) {
1193                 u32 last = lesser(mod(r_q_head + r_q_size),
1194                                   link_last_sent(l_ptr));
1195                 u32 first = buf_seqno(buf);
1196
1197                 while (buf && less(first, r_q_head)) {
1198                         first = mod(first + 1);
1199                         buf = buf->next;
1200                 }
1201                 l_ptr->retransm_queue_head = r_q_head = first;
1202                 l_ptr->retransm_queue_size = r_q_size = mod(last - first);
1203         }
1204
1205         /* Continue retransmission now, if there is anything: */
1206         if (r_q_size && buf) {
1207                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1208                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1209                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1210                 l_ptr->retransm_queue_head = mod(++r_q_head);
1211                 l_ptr->retransm_queue_size = --r_q_size;
1212                 l_ptr->stats.retransmitted++;
1213                 return 0;
1214         }
1215
1216         /* Send deferred protocol message, if any: */
1217         buf = l_ptr->proto_msg_queue;
1218         if (buf) {
1219                 msg_set_ack(buf_msg(buf), mod(l_ptr->next_in_no - 1));
1220                 msg_set_bcast_ack(buf_msg(buf), l_ptr->owner->bclink.last_in);
1221                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1222                 l_ptr->unacked_window = 0;
1223                 kfree_skb(buf);
1224                 l_ptr->proto_msg_queue = NULL;
1225                 return 0;
1226         }
1227
1228         /* Send one deferred data message, if send window not full: */
1229         buf = l_ptr->next_out;
1230         if (buf) {
1231                 struct tipc_msg *msg = buf_msg(buf);
1232                 u32 next = msg_seqno(msg);
1233                 u32 first = buf_seqno(l_ptr->first_out);
1234
1235                 if (mod(next - first) < l_ptr->queue_limit[0]) {
1236                         msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1237                         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1238                         tipc_bearer_send(l_ptr->bearer_id, buf,
1239                                          &l_ptr->media_addr);
1240                         if (msg_user(msg) == MSG_BUNDLER)
1241                                 msg_set_type(msg, CLOSED_MSG);
1242                         l_ptr->next_out = buf->next;
1243                         return 0;
1244                 }
1245         }
1246         return 1;
1247 }
1248
1249 /*
1250  * push_queue(): push out the unsent messages of a link where
1251  *               congestion has abated. Node is locked
1252  */
1253 void tipc_link_push_queue(struct tipc_link *l_ptr)
1254 {
1255         u32 res;
1256
1257         do {
1258                 res = tipc_link_push_packet(l_ptr);
1259         } while (!res);
1260 }
1261
1262 void tipc_link_reset_all(struct tipc_node *node)
1263 {
1264         char addr_string[16];
1265         u32 i;
1266
1267         tipc_node_lock(node);
1268
1269         pr_warn("Resetting all links to %s\n",
1270                 tipc_addr_string_fill(addr_string, node->addr));
1271
1272         for (i = 0; i < MAX_BEARERS; i++) {
1273                 if (node->links[i]) {
1274                         link_print(node->links[i], "Resetting link\n");
1275                         tipc_link_reset(node->links[i]);
1276                 }
1277         }
1278
1279         tipc_node_unlock(node);
1280 }
1281
1282 static void link_retransmit_failure(struct tipc_link *l_ptr,
1283                                     struct sk_buff *buf)
1284 {
1285         struct tipc_msg *msg = buf_msg(buf);
1286
1287         pr_warn("Retransmission failure on link <%s>\n", l_ptr->name);
1288
1289         if (l_ptr->addr) {
1290                 /* Handle failure on standard link */
1291                 link_print(l_ptr, "Resetting link\n");
1292                 tipc_link_reset(l_ptr);
1293
1294         } else {
1295                 /* Handle failure on broadcast link */
1296                 struct tipc_node *n_ptr;
1297                 char addr_string[16];
1298
1299                 pr_info("Msg seq number: %u,  ", msg_seqno(msg));
1300                 pr_cont("Outstanding acks: %lu\n",
1301                         (unsigned long) TIPC_SKB_CB(buf)->handle);
1302
1303                 n_ptr = tipc_bclink_retransmit_to();
1304                 tipc_node_lock(n_ptr);
1305
1306                 tipc_addr_string_fill(addr_string, n_ptr->addr);
1307                 pr_info("Broadcast link info for %s\n", addr_string);
1308                 pr_info("Reception permitted: %d,  Acked: %u\n",
1309                         n_ptr->bclink.recv_permitted,
1310                         n_ptr->bclink.acked);
1311                 pr_info("Last in: %u,  Oos state: %u,  Last sent: %u\n",
1312                         n_ptr->bclink.last_in,
1313                         n_ptr->bclink.oos_state,
1314                         n_ptr->bclink.last_sent);
1315
1316                 tipc_node_unlock(n_ptr);
1317
1318                 tipc_bclink_set_flags(TIPC_BCLINK_RESET);
1319                 l_ptr->stale_count = 0;
1320         }
1321 }
1322
1323 void tipc_link_retransmit(struct tipc_link *l_ptr, struct sk_buff *buf,
1324                           u32 retransmits)
1325 {
1326         struct tipc_msg *msg;
1327
1328         if (!buf)
1329                 return;
1330
1331         msg = buf_msg(buf);
1332
1333         /* Detect repeated retransmit failures */
1334         if (l_ptr->last_retransmitted == msg_seqno(msg)) {
1335                 if (++l_ptr->stale_count > 100) {
1336                         link_retransmit_failure(l_ptr, buf);
1337                         return;
1338                 }
1339         } else {
1340                 l_ptr->last_retransmitted = msg_seqno(msg);
1341                 l_ptr->stale_count = 1;
1342         }
1343
1344         while (retransmits && (buf != l_ptr->next_out) && buf) {
1345                 msg = buf_msg(buf);
1346                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1347                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1348                 tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1349                 buf = buf->next;
1350                 retransmits--;
1351                 l_ptr->stats.retransmitted++;
1352         }
1353
1354         l_ptr->retransm_queue_head = l_ptr->retransm_queue_size = 0;
1355 }
1356
1357 /**
1358  * link_insert_deferred_queue - insert deferred messages back into receive chain
1359  */
1360 static struct sk_buff *link_insert_deferred_queue(struct tipc_link *l_ptr,
1361                                                   struct sk_buff *buf)
1362 {
1363         u32 seq_no;
1364
1365         if (l_ptr->oldest_deferred_in == NULL)
1366                 return buf;
1367
1368         seq_no = buf_seqno(l_ptr->oldest_deferred_in);
1369         if (seq_no == mod(l_ptr->next_in_no)) {
1370                 l_ptr->newest_deferred_in->next = buf;
1371                 buf = l_ptr->oldest_deferred_in;
1372                 l_ptr->oldest_deferred_in = NULL;
1373                 l_ptr->deferred_inqueue_sz = 0;
1374         }
1375         return buf;
1376 }
1377
1378 /**
1379  * link_recv_buf_validate - validate basic format of received message
1380  *
1381  * This routine ensures a TIPC message has an acceptable header, and at least
1382  * as much data as the header indicates it should.  The routine also ensures
1383  * that the entire message header is stored in the main fragment of the message
1384  * buffer, to simplify future access to message header fields.
1385  *
1386  * Note: Having extra info present in the message header or data areas is OK.
1387  * TIPC will ignore the excess, under the assumption that it is optional info
1388  * introduced by a later release of the protocol.
1389  */
1390 static int link_recv_buf_validate(struct sk_buff *buf)
1391 {
1392         static u32 min_data_hdr_size[8] = {
1393                 SHORT_H_SIZE, MCAST_H_SIZE, NAMED_H_SIZE, BASIC_H_SIZE,
1394                 MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE, MAX_H_SIZE
1395                 };
1396
1397         struct tipc_msg *msg;
1398         u32 tipc_hdr[2];
1399         u32 size;
1400         u32 hdr_size;
1401         u32 min_hdr_size;
1402
1403         /* If this packet comes from the defer queue, the skb has already
1404          * been validated
1405          */
1406         if (unlikely(TIPC_SKB_CB(buf)->deferred))
1407                 return 1;
1408
1409         if (unlikely(buf->len < MIN_H_SIZE))
1410                 return 0;
1411
1412         msg = skb_header_pointer(buf, 0, sizeof(tipc_hdr), tipc_hdr);
1413         if (msg == NULL)
1414                 return 0;
1415
1416         if (unlikely(msg_version(msg) != TIPC_VERSION))
1417                 return 0;
1418
1419         size = msg_size(msg);
1420         hdr_size = msg_hdr_sz(msg);
1421         min_hdr_size = msg_isdata(msg) ?
1422                 min_data_hdr_size[msg_type(msg)] : INT_H_SIZE;
1423
1424         if (unlikely((hdr_size < min_hdr_size) ||
1425                      (size < hdr_size) ||
1426                      (buf->len < size) ||
1427                      (size - hdr_size > TIPC_MAX_USER_MSG_SIZE)))
1428                 return 0;
1429
1430         return pskb_may_pull(buf, hdr_size);
1431 }
1432
1433 /**
1434  * tipc_rcv - process TIPC packets/messages arriving from off-node
1435  * @head: pointer to message buffer chain
1436  * @b_ptr: pointer to bearer message arrived on
1437  *
1438  * Invoked with no locks held.  Bearer pointer must point to a valid bearer
1439  * structure (i.e. cannot be NULL), but bearer can be inactive.
1440  */
1441 void tipc_rcv(struct sk_buff *head, struct tipc_bearer *b_ptr)
1442 {
1443         while (head) {
1444                 struct tipc_node *n_ptr;
1445                 struct tipc_link *l_ptr;
1446                 struct sk_buff *crs;
1447                 struct sk_buff *buf = head;
1448                 struct tipc_msg *msg;
1449                 u32 seq_no;
1450                 u32 ackd;
1451                 u32 released = 0;
1452
1453                 head = head->next;
1454                 buf->next = NULL;
1455
1456                 /* Ensure message is well-formed */
1457                 if (unlikely(!link_recv_buf_validate(buf)))
1458                         goto discard;
1459
1460                 /* Ensure message data is a single contiguous unit */
1461                 if (unlikely(skb_linearize(buf)))
1462                         goto discard;
1463
1464                 /* Handle arrival of a non-unicast link message */
1465                 msg = buf_msg(buf);
1466
1467                 if (unlikely(msg_non_seq(msg))) {
1468                         if (msg_user(msg) ==  LINK_CONFIG)
1469                                 tipc_disc_rcv(buf, b_ptr);
1470                         else
1471                                 tipc_bclink_rcv(buf);
1472                         continue;
1473                 }
1474
1475                 /* Discard unicast link messages destined for another node */
1476                 if (unlikely(!msg_short(msg) &&
1477                              (msg_destnode(msg) != tipc_own_addr)))
1478                         goto discard;
1479
1480                 /* Locate neighboring node that sent message */
1481                 n_ptr = tipc_node_find(msg_prevnode(msg));
1482                 if (unlikely(!n_ptr))
1483                         goto discard;
1484                 tipc_node_lock(n_ptr);
1485
1486                 /* Locate unicast link endpoint that should handle message */
1487                 l_ptr = n_ptr->links[b_ptr->identity];
1488                 if (unlikely(!l_ptr))
1489                         goto unlock_discard;
1490
1491                 /* Verify that communication with node is currently allowed */
1492                 if ((n_ptr->action_flags & TIPC_WAIT_PEER_LINKS_DOWN) &&
1493                     msg_user(msg) == LINK_PROTOCOL &&
1494                     (msg_type(msg) == RESET_MSG ||
1495                     msg_type(msg) == ACTIVATE_MSG) &&
1496                     !msg_redundant_link(msg))
1497                         n_ptr->action_flags &= ~TIPC_WAIT_PEER_LINKS_DOWN;
1498
1499                 if (tipc_node_blocked(n_ptr))
1500                         goto unlock_discard;
1501
1502                 /* Validate message sequence number info */
1503                 seq_no = msg_seqno(msg);
1504                 ackd = msg_ack(msg);
1505
1506                 /* Release acked messages */
1507                 if (n_ptr->bclink.recv_permitted)
1508                         tipc_bclink_acknowledge(n_ptr, msg_bcast_ack(msg));
1509
1510                 crs = l_ptr->first_out;
1511                 while ((crs != l_ptr->next_out) &&
1512                        less_eq(buf_seqno(crs), ackd)) {
1513                         struct sk_buff *next = crs->next;
1514                         kfree_skb(crs);
1515                         crs = next;
1516                         released++;
1517                 }
1518                 if (released) {
1519                         l_ptr->first_out = crs;
1520                         l_ptr->out_queue_size -= released;
1521                 }
1522
1523                 /* Try sending any messages link endpoint has pending */
1524                 if (unlikely(l_ptr->next_out))
1525                         tipc_link_push_queue(l_ptr);
1526
1527                 if (unlikely(!list_empty(&l_ptr->waiting_ports)))
1528                         tipc_link_wakeup_ports(l_ptr, 0);
1529
1530                 if (unlikely(++l_ptr->unacked_window >= TIPC_MIN_LINK_WIN)) {
1531                         l_ptr->stats.sent_acks++;
1532                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1533                 }
1534
1535                 /* Process the incoming packet */
1536                 if (unlikely(!link_working_working(l_ptr))) {
1537                         if (msg_user(msg) == LINK_PROTOCOL) {
1538                                 tipc_link_proto_rcv(l_ptr, buf);
1539                                 head = link_insert_deferred_queue(l_ptr, head);
1540                                 tipc_node_unlock(n_ptr);
1541                                 continue;
1542                         }
1543
1544                         /* Traffic message. Conditionally activate link */
1545                         link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1546
1547                         if (link_working_working(l_ptr)) {
1548                                 /* Re-insert buffer in front of queue */
1549                                 buf->next = head;
1550                                 head = buf;
1551                                 tipc_node_unlock(n_ptr);
1552                                 continue;
1553                         }
1554                         goto unlock_discard;
1555                 }
1556
1557                 /* Link is now in state WORKING_WORKING */
1558                 if (unlikely(seq_no != mod(l_ptr->next_in_no))) {
1559                         link_handle_out_of_seq_msg(l_ptr, buf);
1560                         head = link_insert_deferred_queue(l_ptr, head);
1561                         tipc_node_unlock(n_ptr);
1562                         continue;
1563                 }
1564                 l_ptr->next_in_no++;
1565                 if (unlikely(l_ptr->oldest_deferred_in))
1566                         head = link_insert_deferred_queue(l_ptr, head);
1567
1568                 /* Deliver packet/message to correct user: */
1569                 if (unlikely(msg_user(msg) ==  CHANGEOVER_PROTOCOL)) {
1570                         if (!tipc_link_tunnel_rcv(n_ptr, &buf)) {
1571                                 tipc_node_unlock(n_ptr);
1572                                 continue;
1573                         }
1574                         msg = buf_msg(buf);
1575                 } else if (msg_user(msg) == MSG_FRAGMENTER) {
1576                         int rc;
1577
1578                         l_ptr->stats.recv_fragments++;
1579                         rc = tipc_link_frag_rcv(&l_ptr->reasm_head,
1580                                                 &l_ptr->reasm_tail,
1581                                                 &buf);
1582                         if (rc == LINK_REASM_COMPLETE) {
1583                                 l_ptr->stats.recv_fragmented++;
1584                                 msg = buf_msg(buf);
1585                         } else {
1586                                 if (rc == LINK_REASM_ERROR)
1587                                         tipc_link_reset(l_ptr);
1588                                 tipc_node_unlock(n_ptr);
1589                                 continue;
1590                         }
1591                 }
1592
1593                 switch (msg_user(msg)) {
1594                 case TIPC_LOW_IMPORTANCE:
1595                 case TIPC_MEDIUM_IMPORTANCE:
1596                 case TIPC_HIGH_IMPORTANCE:
1597                 case TIPC_CRITICAL_IMPORTANCE:
1598                         tipc_node_unlock(n_ptr);
1599                         tipc_port_rcv(buf);
1600                         continue;
1601                 case MSG_BUNDLER:
1602                         l_ptr->stats.recv_bundles++;
1603                         l_ptr->stats.recv_bundled += msg_msgcnt(msg);
1604                         tipc_node_unlock(n_ptr);
1605                         tipc_link_bundle_rcv(buf);
1606                         continue;
1607                 case NAME_DISTRIBUTOR:
1608                         n_ptr->bclink.recv_permitted = true;
1609                         tipc_node_unlock(n_ptr);
1610                         tipc_named_rcv(buf);
1611                         continue;
1612                 case CONN_MANAGER:
1613                         tipc_node_unlock(n_ptr);
1614                         tipc_port_proto_rcv(buf);
1615                         continue;
1616                 case BCAST_PROTOCOL:
1617                         tipc_link_sync_rcv(n_ptr, buf);
1618                         break;
1619                 default:
1620                         kfree_skb(buf);
1621                         break;
1622                 }
1623                 tipc_node_unlock(n_ptr);
1624                 continue;
1625 unlock_discard:
1626                 tipc_node_unlock(n_ptr);
1627 discard:
1628                 kfree_skb(buf);
1629         }
1630 }
1631
1632 /**
1633  * tipc_link_defer_pkt - Add out-of-sequence message to deferred reception queue
1634  *
1635  * Returns increase in queue length (i.e. 0 or 1)
1636  */
1637 u32 tipc_link_defer_pkt(struct sk_buff **head, struct sk_buff **tail,
1638                         struct sk_buff *buf)
1639 {
1640         struct sk_buff *queue_buf;
1641         struct sk_buff **prev;
1642         u32 seq_no = buf_seqno(buf);
1643
1644         buf->next = NULL;
1645
1646         /* Empty queue ? */
1647         if (*head == NULL) {
1648                 *head = *tail = buf;
1649                 return 1;
1650         }
1651
1652         /* Last ? */
1653         if (less(buf_seqno(*tail), seq_no)) {
1654                 (*tail)->next = buf;
1655                 *tail = buf;
1656                 return 1;
1657         }
1658
1659         /* Locate insertion point in queue, then insert; discard if duplicate */
1660         prev = head;
1661         queue_buf = *head;
1662         for (;;) {
1663                 u32 curr_seqno = buf_seqno(queue_buf);
1664
1665                 if (seq_no == curr_seqno) {
1666                         kfree_skb(buf);
1667                         return 0;
1668                 }
1669
1670                 if (less(seq_no, curr_seqno))
1671                         break;
1672
1673                 prev = &queue_buf->next;
1674                 queue_buf = queue_buf->next;
1675         }
1676
1677         buf->next = queue_buf;
1678         *prev = buf;
1679         return 1;
1680 }
1681
1682 /*
1683  * link_handle_out_of_seq_msg - handle arrival of out-of-sequence packet
1684  */
1685 static void link_handle_out_of_seq_msg(struct tipc_link *l_ptr,
1686                                        struct sk_buff *buf)
1687 {
1688         u32 seq_no = buf_seqno(buf);
1689
1690         if (likely(msg_user(buf_msg(buf)) == LINK_PROTOCOL)) {
1691                 tipc_link_proto_rcv(l_ptr, buf);
1692                 return;
1693         }
1694
1695         /* Record OOS packet arrival (force mismatch on next timeout) */
1696         l_ptr->checkpoint--;
1697
1698         /*
1699          * Discard packet if a duplicate; otherwise add it to deferred queue
1700          * and notify peer of gap as per protocol specification
1701          */
1702         if (less(seq_no, mod(l_ptr->next_in_no))) {
1703                 l_ptr->stats.duplicates++;
1704                 kfree_skb(buf);
1705                 return;
1706         }
1707
1708         if (tipc_link_defer_pkt(&l_ptr->oldest_deferred_in,
1709                                 &l_ptr->newest_deferred_in, buf)) {
1710                 l_ptr->deferred_inqueue_sz++;
1711                 l_ptr->stats.deferred_recv++;
1712                 TIPC_SKB_CB(buf)->deferred = true;
1713                 if ((l_ptr->deferred_inqueue_sz % 16) == 1)
1714                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0, 0, 0, 0);
1715         } else
1716                 l_ptr->stats.duplicates++;
1717 }
1718
1719 /*
1720  * Send protocol message to the other endpoint.
1721  */
1722 void tipc_link_proto_xmit(struct tipc_link *l_ptr, u32 msg_typ, int probe_msg,
1723                           u32 gap, u32 tolerance, u32 priority, u32 ack_mtu)
1724 {
1725         struct sk_buff *buf = NULL;
1726         struct tipc_msg *msg = l_ptr->pmsg;
1727         u32 msg_size = sizeof(l_ptr->proto_msg);
1728         int r_flag;
1729
1730         /* Discard any previous message that was deferred due to congestion */
1731         if (l_ptr->proto_msg_queue) {
1732                 kfree_skb(l_ptr->proto_msg_queue);
1733                 l_ptr->proto_msg_queue = NULL;
1734         }
1735
1736         /* Don't send protocol message during link changeover */
1737         if (l_ptr->exp_msg_count)
1738                 return;
1739
1740         /* Abort non-RESET send if communication with node is prohibited */
1741         if ((tipc_node_blocked(l_ptr->owner)) && (msg_typ != RESET_MSG))
1742                 return;
1743
1744         /* Create protocol message with "out-of-sequence" sequence number */
1745         msg_set_type(msg, msg_typ);
1746         msg_set_net_plane(msg, l_ptr->net_plane);
1747         msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
1748         msg_set_last_bcast(msg, tipc_bclink_get_last_sent());
1749
1750         if (msg_typ == STATE_MSG) {
1751                 u32 next_sent = mod(l_ptr->next_out_no);
1752
1753                 if (!tipc_link_is_up(l_ptr))
1754                         return;
1755                 if (l_ptr->next_out)
1756                         next_sent = buf_seqno(l_ptr->next_out);
1757                 msg_set_next_sent(msg, next_sent);
1758                 if (l_ptr->oldest_deferred_in) {
1759                         u32 rec = buf_seqno(l_ptr->oldest_deferred_in);
1760                         gap = mod(rec - mod(l_ptr->next_in_no));
1761                 }
1762                 msg_set_seq_gap(msg, gap);
1763                 if (gap)
1764                         l_ptr->stats.sent_nacks++;
1765                 msg_set_link_tolerance(msg, tolerance);
1766                 msg_set_linkprio(msg, priority);
1767                 msg_set_max_pkt(msg, ack_mtu);
1768                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));
1769                 msg_set_probe(msg, probe_msg != 0);
1770                 if (probe_msg) {
1771                         u32 mtu = l_ptr->max_pkt;
1772
1773                         if ((mtu < l_ptr->max_pkt_target) &&
1774                             link_working_working(l_ptr) &&
1775                             l_ptr->fsm_msg_cnt) {
1776                                 msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1777                                 if (l_ptr->max_pkt_probes == 10) {
1778                                         l_ptr->max_pkt_target = (msg_size - 4);
1779                                         l_ptr->max_pkt_probes = 0;
1780                                         msg_size = (mtu + (l_ptr->max_pkt_target - mtu)/2 + 2) & ~3;
1781                                 }
1782                                 l_ptr->max_pkt_probes++;
1783                         }
1784
1785                         l_ptr->stats.sent_probes++;
1786                 }
1787                 l_ptr->stats.sent_states++;
1788         } else {                /* RESET_MSG or ACTIVATE_MSG */
1789                 msg_set_ack(msg, mod(l_ptr->reset_checkpoint - 1));
1790                 msg_set_seq_gap(msg, 0);
1791                 msg_set_next_sent(msg, 1);
1792                 msg_set_probe(msg, 0);
1793                 msg_set_link_tolerance(msg, l_ptr->tolerance);
1794                 msg_set_linkprio(msg, l_ptr->priority);
1795                 msg_set_max_pkt(msg, l_ptr->max_pkt_target);
1796         }
1797
1798         r_flag = (l_ptr->owner->working_links > tipc_link_is_up(l_ptr));
1799         msg_set_redundant_link(msg, r_flag);
1800         msg_set_linkprio(msg, l_ptr->priority);
1801         msg_set_size(msg, msg_size);
1802
1803         msg_set_seqno(msg, mod(l_ptr->next_out_no + (0xffff/2)));
1804
1805         buf = tipc_buf_acquire(msg_size);
1806         if (!buf)
1807                 return;
1808
1809         skb_copy_to_linear_data(buf, msg, sizeof(l_ptr->proto_msg));
1810         buf->priority = TC_PRIO_CONTROL;
1811
1812         tipc_bearer_send(l_ptr->bearer_id, buf, &l_ptr->media_addr);
1813         l_ptr->unacked_window = 0;
1814         kfree_skb(buf);
1815 }
1816
1817 /*
1818  * Receive protocol message :
1819  * Note that network plane id propagates through the network, and may
1820  * change at any time. The node with lowest address rules
1821  */
1822 static void tipc_link_proto_rcv(struct tipc_link *l_ptr, struct sk_buff *buf)
1823 {
1824         u32 rec_gap = 0;
1825         u32 max_pkt_info;
1826         u32 max_pkt_ack;
1827         u32 msg_tol;
1828         struct tipc_msg *msg = buf_msg(buf);
1829
1830         /* Discard protocol message during link changeover */
1831         if (l_ptr->exp_msg_count)
1832                 goto exit;
1833
1834         if (l_ptr->net_plane != msg_net_plane(msg))
1835                 if (tipc_own_addr > msg_prevnode(msg))
1836                         l_ptr->net_plane = msg_net_plane(msg);
1837
1838         switch (msg_type(msg)) {
1839
1840         case RESET_MSG:
1841                 if (!link_working_unknown(l_ptr) &&
1842                     (l_ptr->peer_session != INVALID_SESSION)) {
1843                         if (less_eq(msg_session(msg), l_ptr->peer_session))
1844                                 break; /* duplicate or old reset: ignore */
1845                 }
1846
1847                 if (!msg_redundant_link(msg) && (link_working_working(l_ptr) ||
1848                                 link_working_unknown(l_ptr))) {
1849                         /*
1850                          * peer has lost contact -- don't allow peer's links
1851                          * to reactivate before we recognize loss & clean up
1852                          */
1853                         l_ptr->owner->action_flags |= TIPC_WAIT_OWN_LINKS_DOWN;
1854                 }
1855
1856                 link_state_event(l_ptr, RESET_MSG);
1857
1858                 /* fall thru' */
1859         case ACTIVATE_MSG:
1860                 /* Update link settings according other endpoint's values */
1861                 strcpy((strrchr(l_ptr->name, ':') + 1), (char *)msg_data(msg));
1862
1863                 msg_tol = msg_link_tolerance(msg);
1864                 if (msg_tol > l_ptr->tolerance)
1865                         link_set_supervision_props(l_ptr, msg_tol);
1866
1867                 if (msg_linkprio(msg) > l_ptr->priority)
1868                         l_ptr->priority = msg_linkprio(msg);
1869
1870                 max_pkt_info = msg_max_pkt(msg);
1871                 if (max_pkt_info) {
1872                         if (max_pkt_info < l_ptr->max_pkt_target)
1873                                 l_ptr->max_pkt_target = max_pkt_info;
1874                         if (l_ptr->max_pkt > l_ptr->max_pkt_target)
1875                                 l_ptr->max_pkt = l_ptr->max_pkt_target;
1876                 } else {
1877                         l_ptr->max_pkt = l_ptr->max_pkt_target;
1878                 }
1879
1880                 /* Synchronize broadcast link info, if not done previously */
1881                 if (!tipc_node_is_up(l_ptr->owner)) {
1882                         l_ptr->owner->bclink.last_sent =
1883                                 l_ptr->owner->bclink.last_in =
1884                                 msg_last_bcast(msg);
1885                         l_ptr->owner->bclink.oos_state = 0;
1886                 }
1887
1888                 l_ptr->peer_session = msg_session(msg);
1889                 l_ptr->peer_bearer_id = msg_bearer_id(msg);
1890
1891                 if (msg_type(msg) == ACTIVATE_MSG)
1892                         link_state_event(l_ptr, ACTIVATE_MSG);
1893                 break;
1894         case STATE_MSG:
1895
1896                 msg_tol = msg_link_tolerance(msg);
1897                 if (msg_tol)
1898                         link_set_supervision_props(l_ptr, msg_tol);
1899
1900                 if (msg_linkprio(msg) &&
1901                     (msg_linkprio(msg) != l_ptr->priority)) {
1902                         pr_warn("%s<%s>, priority change %u->%u\n",
1903                                 link_rst_msg, l_ptr->name, l_ptr->priority,
1904                                 msg_linkprio(msg));
1905                         l_ptr->priority = msg_linkprio(msg);
1906                         tipc_link_reset(l_ptr); /* Enforce change to take effect */
1907                         break;
1908                 }
1909
1910                 /* Record reception; force mismatch at next timeout: */
1911                 l_ptr->checkpoint--;
1912
1913                 link_state_event(l_ptr, TRAFFIC_MSG_EVT);
1914                 l_ptr->stats.recv_states++;
1915                 if (link_reset_unknown(l_ptr))
1916                         break;
1917
1918                 if (less_eq(mod(l_ptr->next_in_no), msg_next_sent(msg))) {
1919                         rec_gap = mod(msg_next_sent(msg) -
1920                                       mod(l_ptr->next_in_no));
1921                 }
1922
1923                 max_pkt_ack = msg_max_pkt(msg);
1924                 if (max_pkt_ack > l_ptr->max_pkt) {
1925                         l_ptr->max_pkt = max_pkt_ack;
1926                         l_ptr->max_pkt_probes = 0;
1927                 }
1928
1929                 max_pkt_ack = 0;
1930                 if (msg_probe(msg)) {
1931                         l_ptr->stats.recv_probes++;
1932                         if (msg_size(msg) > sizeof(l_ptr->proto_msg))
1933                                 max_pkt_ack = msg_size(msg);
1934                 }
1935
1936                 /* Protocol message before retransmits, reduce loss risk */
1937                 if (l_ptr->owner->bclink.recv_permitted)
1938                         tipc_bclink_update_link_state(l_ptr->owner,
1939                                                       msg_last_bcast(msg));
1940
1941                 if (rec_gap || (msg_probe(msg))) {
1942                         tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, rec_gap, 0,
1943                                              0, max_pkt_ack);
1944                 }
1945                 if (msg_seq_gap(msg)) {
1946                         l_ptr->stats.recv_nacks++;
1947                         tipc_link_retransmit(l_ptr, l_ptr->first_out,
1948                                              msg_seq_gap(msg));
1949                 }
1950                 break;
1951         }
1952 exit:
1953         kfree_skb(buf);
1954 }
1955
1956
1957 /* tipc_link_tunnel_xmit(): Tunnel one packet via a link belonging to
1958  * a different bearer. Owner node is locked.
1959  */
1960 static void tipc_link_tunnel_xmit(struct tipc_link *l_ptr,
1961                                   struct tipc_msg *tunnel_hdr,
1962                                   struct tipc_msg *msg,
1963                                   u32 selector)
1964 {
1965         struct tipc_link *tunnel;
1966         struct sk_buff *buf;
1967         u32 length = msg_size(msg);
1968
1969         tunnel = l_ptr->owner->active_links[selector & 1];
1970         if (!tipc_link_is_up(tunnel)) {
1971                 pr_warn("%stunnel link no longer available\n", link_co_err);
1972                 return;
1973         }
1974         msg_set_size(tunnel_hdr, length + INT_H_SIZE);
1975         buf = tipc_buf_acquire(length + INT_H_SIZE);
1976         if (!buf) {
1977                 pr_warn("%sunable to send tunnel msg\n", link_co_err);
1978                 return;
1979         }
1980         skb_copy_to_linear_data(buf, tunnel_hdr, INT_H_SIZE);
1981         skb_copy_to_linear_data_offset(buf, INT_H_SIZE, msg, length);
1982         __tipc_link_xmit(tunnel, buf);
1983 }
1984
1985
1986 /* tipc_link_failover_send_queue(): A link has gone down, but a second
1987  * link is still active. We can do failover. Tunnel the failing link's
1988  * whole send queue via the remaining link. This way, we don't lose
1989  * any packets, and sequence order is preserved for subsequent traffic
1990  * sent over the remaining link. Owner node is locked.
1991  */
1992 void tipc_link_failover_send_queue(struct tipc_link *l_ptr)
1993 {
1994         u32 msgcount = l_ptr->out_queue_size;
1995         struct sk_buff *crs = l_ptr->first_out;
1996         struct tipc_link *tunnel = l_ptr->owner->active_links[0];
1997         struct tipc_msg tunnel_hdr;
1998         int split_bundles;
1999
2000         if (!tunnel)
2001                 return;
2002
2003         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2004                  ORIGINAL_MSG, INT_H_SIZE, l_ptr->addr);
2005         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2006         msg_set_msgcnt(&tunnel_hdr, msgcount);
2007
2008         if (!l_ptr->first_out) {
2009                 struct sk_buff *buf;
2010
2011                 buf = tipc_buf_acquire(INT_H_SIZE);
2012                 if (buf) {
2013                         skb_copy_to_linear_data(buf, &tunnel_hdr, INT_H_SIZE);
2014                         msg_set_size(&tunnel_hdr, INT_H_SIZE);
2015                         __tipc_link_xmit(tunnel, buf);
2016                 } else {
2017                         pr_warn("%sunable to send changeover msg\n",
2018                                 link_co_err);
2019                 }
2020                 return;
2021         }
2022
2023         split_bundles = (l_ptr->owner->active_links[0] !=
2024                          l_ptr->owner->active_links[1]);
2025
2026         while (crs) {
2027                 struct tipc_msg *msg = buf_msg(crs);
2028
2029                 if ((msg_user(msg) == MSG_BUNDLER) && split_bundles) {
2030                         struct tipc_msg *m = msg_get_wrapped(msg);
2031                         unchar *pos = (unchar *)m;
2032
2033                         msgcount = msg_msgcnt(msg);
2034                         while (msgcount--) {
2035                                 msg_set_seqno(m, msg_seqno(msg));
2036                                 tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, m,
2037                                                       msg_link_selector(m));
2038                                 pos += align(msg_size(m));
2039                                 m = (struct tipc_msg *)pos;
2040                         }
2041                 } else {
2042                         tipc_link_tunnel_xmit(l_ptr, &tunnel_hdr, msg,
2043                                               msg_link_selector(msg));
2044                 }
2045                 crs = crs->next;
2046         }
2047 }
2048
2049 /* tipc_link_dup_queue_xmit(): A second link has become active. Tunnel a
2050  * duplicate of the first link's send queue via the new link. This way, we
2051  * are guaranteed that currently queued packets from a socket are delivered
2052  * before future traffic from the same socket, even if this is using the
2053  * new link. The last arriving copy of each duplicate packet is dropped at
2054  * the receiving end by the regular protocol check, so packet cardinality
2055  * and sequence order is preserved per sender/receiver socket pair.
2056  * Owner node is locked.
2057  */
2058 void tipc_link_dup_queue_xmit(struct tipc_link *l_ptr,
2059                               struct tipc_link *tunnel)
2060 {
2061         struct sk_buff *iter;
2062         struct tipc_msg tunnel_hdr;
2063
2064         tipc_msg_init(&tunnel_hdr, CHANGEOVER_PROTOCOL,
2065                  DUPLICATE_MSG, INT_H_SIZE, l_ptr->addr);
2066         msg_set_msgcnt(&tunnel_hdr, l_ptr->out_queue_size);
2067         msg_set_bearer_id(&tunnel_hdr, l_ptr->peer_bearer_id);
2068         iter = l_ptr->first_out;
2069         while (iter) {
2070                 struct sk_buff *outbuf;
2071                 struct tipc_msg *msg = buf_msg(iter);
2072                 u32 length = msg_size(msg);
2073
2074                 if (msg_user(msg) == MSG_BUNDLER)
2075                         msg_set_type(msg, CLOSED_MSG);
2076                 msg_set_ack(msg, mod(l_ptr->next_in_no - 1));   /* Update */
2077                 msg_set_bcast_ack(msg, l_ptr->owner->bclink.last_in);
2078                 msg_set_size(&tunnel_hdr, length + INT_H_SIZE);
2079                 outbuf = tipc_buf_acquire(length + INT_H_SIZE);
2080                 if (outbuf == NULL) {
2081                         pr_warn("%sunable to send duplicate msg\n",
2082                                 link_co_err);
2083                         return;
2084                 }
2085                 skb_copy_to_linear_data(outbuf, &tunnel_hdr, INT_H_SIZE);
2086                 skb_copy_to_linear_data_offset(outbuf, INT_H_SIZE, iter->data,
2087                                                length);
2088                 __tipc_link_xmit(tunnel, outbuf);
2089                 if (!tipc_link_is_up(l_ptr))
2090                         return;
2091                 iter = iter->next;
2092         }
2093 }
2094
2095 /**
2096  * buf_extract - extracts embedded TIPC message from another message
2097  * @skb: encapsulating message buffer
2098  * @from_pos: offset to extract from
2099  *
2100  * Returns a new message buffer containing an embedded message.  The
2101  * encapsulating message itself is left unchanged.
2102  */
2103 static struct sk_buff *buf_extract(struct sk_buff *skb, u32 from_pos)
2104 {
2105         struct tipc_msg *msg = (struct tipc_msg *)(skb->data + from_pos);
2106         u32 size = msg_size(msg);
2107         struct sk_buff *eb;
2108
2109         eb = tipc_buf_acquire(size);
2110         if (eb)
2111                 skb_copy_to_linear_data(eb, msg, size);
2112         return eb;
2113 }
2114
2115
2116
2117 /* tipc_link_dup_rcv(): Receive a tunnelled DUPLICATE_MSG packet.
2118  * Owner node is locked.
2119  */
2120 static void tipc_link_dup_rcv(struct tipc_link *l_ptr,
2121                               struct sk_buff *t_buf)
2122 {
2123         struct sk_buff *buf;
2124
2125         if (!tipc_link_is_up(l_ptr))
2126                 return;
2127
2128         buf = buf_extract(t_buf, INT_H_SIZE);
2129         if (buf == NULL) {
2130                 pr_warn("%sfailed to extract inner dup pkt\n", link_co_err);
2131                 return;
2132         }
2133
2134         /* Add buffer to deferred queue, if applicable: */
2135         link_handle_out_of_seq_msg(l_ptr, buf);
2136 }
2137
2138 /*  tipc_link_failover_rcv(): Receive a tunnelled ORIGINAL_MSG packet
2139  *  Owner node is locked.
2140  */
2141 static struct sk_buff *tipc_link_failover_rcv(struct tipc_link *l_ptr,
2142                                               struct sk_buff *t_buf)
2143 {
2144         struct tipc_msg *t_msg = buf_msg(t_buf);
2145         struct sk_buff *buf = NULL;
2146         struct tipc_msg *msg;
2147
2148         if (tipc_link_is_up(l_ptr))
2149                 tipc_link_reset(l_ptr);
2150
2151         /* First failover packet? */
2152         if (l_ptr->exp_msg_count == START_CHANGEOVER)
2153                 l_ptr->exp_msg_count = msg_msgcnt(t_msg);
2154
2155         /* Should there be an inner packet? */
2156         if (l_ptr->exp_msg_count) {
2157                 l_ptr->exp_msg_count--;
2158                 buf = buf_extract(t_buf, INT_H_SIZE);
2159                 if (buf == NULL) {
2160                         pr_warn("%sno inner failover pkt\n", link_co_err);
2161                         goto exit;
2162                 }
2163                 msg = buf_msg(buf);
2164
2165                 if (less(msg_seqno(msg), l_ptr->reset_checkpoint)) {
2166                         kfree_skb(buf);
2167                         buf = NULL;
2168                         goto exit;
2169                 }
2170                 if (msg_user(msg) == MSG_FRAGMENTER) {
2171                         l_ptr->stats.recv_fragments++;
2172                         tipc_link_frag_rcv(&l_ptr->reasm_head,
2173                                            &l_ptr->reasm_tail,
2174                                            &buf);
2175                 }
2176         }
2177 exit:
2178         if ((l_ptr->exp_msg_count == 0) && (l_ptr->flags & LINK_STOPPED)) {
2179                 tipc_node_detach_link(l_ptr->owner, l_ptr);
2180                 kfree(l_ptr);
2181         }
2182         return buf;
2183 }
2184
2185 /*  tipc_link_tunnel_rcv(): Receive a tunnelled packet, sent
2186  *  via other link as result of a failover (ORIGINAL_MSG) or
2187  *  a new active link (DUPLICATE_MSG). Failover packets are
2188  *  returned to the active link for delivery upwards.
2189  *  Owner node is locked.
2190  */
2191 static int tipc_link_tunnel_rcv(struct tipc_node *n_ptr,
2192                                 struct sk_buff **buf)
2193 {
2194         struct sk_buff *t_buf = *buf;
2195         struct tipc_link *l_ptr;
2196         struct tipc_msg *t_msg = buf_msg(t_buf);
2197         u32 bearer_id = msg_bearer_id(t_msg);
2198
2199         *buf = NULL;
2200
2201         if (bearer_id >= MAX_BEARERS)
2202                 goto exit;
2203
2204         l_ptr = n_ptr->links[bearer_id];
2205         if (!l_ptr)
2206                 goto exit;
2207
2208         if (msg_type(t_msg) == DUPLICATE_MSG)
2209                 tipc_link_dup_rcv(l_ptr, t_buf);
2210         else if (msg_type(t_msg) == ORIGINAL_MSG)
2211                 *buf = tipc_link_failover_rcv(l_ptr, t_buf);
2212         else
2213                 pr_warn("%sunknown tunnel pkt received\n", link_co_err);
2214 exit:
2215         kfree_skb(t_buf);
2216         return *buf != NULL;
2217 }
2218
2219 /*
2220  *  Bundler functionality:
2221  */
2222 void tipc_link_bundle_rcv(struct sk_buff *buf)
2223 {
2224         u32 msgcount = msg_msgcnt(buf_msg(buf));
2225         u32 pos = INT_H_SIZE;
2226         struct sk_buff *obuf;
2227
2228         while (msgcount--) {
2229                 obuf = buf_extract(buf, pos);
2230                 if (obuf == NULL) {
2231                         pr_warn("Link unable to unbundle message(s)\n");
2232                         break;
2233                 }
2234                 pos += align(msg_size(buf_msg(obuf)));
2235                 tipc_net_route_msg(obuf);
2236         }
2237         kfree_skb(buf);
2238 }
2239
2240 /*
2241  *  Fragmentation/defragmentation:
2242  */
2243
2244 /*
2245  * tipc_link_frag_xmit: Entry for buffers needing fragmentation.
2246  * The buffer is complete, inclusive total message length.
2247  * Returns user data length.
2248  */
2249 static int tipc_link_frag_xmit(struct tipc_link *l_ptr, struct sk_buff *buf)
2250 {
2251         struct sk_buff *buf_chain = NULL;
2252         struct sk_buff *buf_chain_tail = (struct sk_buff *)&buf_chain;
2253         struct tipc_msg *inmsg = buf_msg(buf);
2254         struct tipc_msg fragm_hdr;
2255         u32 insize = msg_size(inmsg);
2256         u32 dsz = msg_data_sz(inmsg);
2257         unchar *crs = buf->data;
2258         u32 rest = insize;
2259         u32 pack_sz = l_ptr->max_pkt;
2260         u32 fragm_sz = pack_sz - INT_H_SIZE;
2261         u32 fragm_no = 0;
2262         u32 destaddr;
2263
2264         if (msg_short(inmsg))
2265                 destaddr = l_ptr->addr;
2266         else
2267                 destaddr = msg_destnode(inmsg);
2268
2269         /* Prepare reusable fragment header: */
2270         tipc_msg_init(&fragm_hdr, MSG_FRAGMENTER, FIRST_FRAGMENT,
2271                  INT_H_SIZE, destaddr);
2272
2273         /* Chop up message: */
2274         while (rest > 0) {
2275                 struct sk_buff *fragm;
2276
2277                 if (rest <= fragm_sz) {
2278                         fragm_sz = rest;
2279                         msg_set_type(&fragm_hdr, LAST_FRAGMENT);
2280                 }
2281                 fragm = tipc_buf_acquire(fragm_sz + INT_H_SIZE);
2282                 if (fragm == NULL) {
2283                         kfree_skb(buf);
2284                         kfree_skb_list(buf_chain);
2285                         return -ENOMEM;
2286                 }
2287                 msg_set_size(&fragm_hdr, fragm_sz + INT_H_SIZE);
2288                 fragm_no++;
2289                 msg_set_fragm_no(&fragm_hdr, fragm_no);
2290                 skb_copy_to_linear_data(fragm, &fragm_hdr, INT_H_SIZE);
2291                 skb_copy_to_linear_data_offset(fragm, INT_H_SIZE, crs,
2292                                                fragm_sz);
2293                 buf_chain_tail->next = fragm;
2294                 buf_chain_tail = fragm;
2295
2296                 rest -= fragm_sz;
2297                 crs += fragm_sz;
2298                 msg_set_type(&fragm_hdr, FRAGMENT);
2299         }
2300         kfree_skb(buf);
2301
2302         /* Append chain of fragments to send queue & send them */
2303         l_ptr->long_msg_seq_no++;
2304         link_add_chain_to_outqueue(l_ptr, buf_chain, l_ptr->long_msg_seq_no);
2305         l_ptr->stats.sent_fragments += fragm_no;
2306         l_ptr->stats.sent_fragmented++;
2307         tipc_link_push_queue(l_ptr);
2308
2309         return dsz;
2310 }
2311
2312 /* tipc_link_frag_rcv(): Called with node lock on. Returns
2313  * the reassembled buffer if message is complete.
2314  */
2315 int tipc_link_frag_rcv(struct sk_buff **head, struct sk_buff **tail,
2316                        struct sk_buff **fbuf)
2317 {
2318         struct sk_buff *frag = *fbuf;
2319         struct tipc_msg *msg = buf_msg(frag);
2320         u32 fragid = msg_type(msg);
2321         bool headstolen;
2322         int delta;
2323
2324         skb_pull(frag, msg_hdr_sz(msg));
2325         if (fragid == FIRST_FRAGMENT) {
2326                 if (*head || skb_unclone(frag, GFP_ATOMIC))
2327                         goto out_free;
2328                 *head = frag;
2329                 skb_frag_list_init(*head);
2330                 *fbuf = NULL;
2331                 return 0;
2332         } else if (*head &&
2333                    skb_try_coalesce(*head, frag, &headstolen, &delta)) {
2334                 kfree_skb_partial(frag, headstolen);
2335         } else {
2336                 if (!*head)
2337                         goto out_free;
2338                 if (!skb_has_frag_list(*head))
2339                         skb_shinfo(*head)->frag_list = frag;
2340                 else
2341                         (*tail)->next = frag;
2342                 *tail = frag;
2343                 (*head)->truesize += frag->truesize;
2344         }
2345         if (fragid == LAST_FRAGMENT) {
2346                 *fbuf = *head;
2347                 *tail = *head = NULL;
2348                 return LINK_REASM_COMPLETE;
2349         }
2350         *fbuf = NULL;
2351         return 0;
2352 out_free:
2353         pr_warn_ratelimited("Link unable to reassemble fragmented message\n");
2354         kfree_skb(*fbuf);
2355         *fbuf = NULL;
2356         return LINK_REASM_ERROR;
2357 }
2358
2359 static void link_set_supervision_props(struct tipc_link *l_ptr, u32 tolerance)
2360 {
2361         if ((tolerance < TIPC_MIN_LINK_TOL) || (tolerance > TIPC_MAX_LINK_TOL))
2362                 return;
2363
2364         l_ptr->tolerance = tolerance;
2365         l_ptr->continuity_interval =
2366                 ((tolerance / 4) > 500) ? 500 : tolerance / 4;
2367         l_ptr->abort_limit = tolerance / (l_ptr->continuity_interval / 4);
2368 }
2369
2370 void tipc_link_set_queue_limits(struct tipc_link *l_ptr, u32 window)
2371 {
2372         /* Data messages from this node, inclusive FIRST_FRAGM */
2373         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE] = window;
2374         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE] = (window / 3) * 4;
2375         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE] = (window / 3) * 5;
2376         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE] = (window / 3) * 6;
2377         /* Transiting data messages,inclusive FIRST_FRAGM */
2378         l_ptr->queue_limit[TIPC_LOW_IMPORTANCE + 4] = 300;
2379         l_ptr->queue_limit[TIPC_MEDIUM_IMPORTANCE + 4] = 600;
2380         l_ptr->queue_limit[TIPC_HIGH_IMPORTANCE + 4] = 900;
2381         l_ptr->queue_limit[TIPC_CRITICAL_IMPORTANCE + 4] = 1200;
2382         l_ptr->queue_limit[CONN_MANAGER] = 1200;
2383         l_ptr->queue_limit[CHANGEOVER_PROTOCOL] = 2500;
2384         l_ptr->queue_limit[NAME_DISTRIBUTOR] = 3000;
2385         /* FRAGMENT and LAST_FRAGMENT packets */
2386         l_ptr->queue_limit[MSG_FRAGMENTER] = 4000;
2387 }
2388
2389 /* tipc_link_find_owner - locate owner node of link by link's name
2390  * @name: pointer to link name string
2391  * @bearer_id: pointer to index in 'node->links' array where the link was found.
2392  *
2393  * Returns pointer to node owning the link, or 0 if no matching link is found.
2394  */
2395 static struct tipc_node *tipc_link_find_owner(const char *link_name,
2396                                               unsigned int *bearer_id)
2397 {
2398         struct tipc_link *l_ptr;
2399         struct tipc_node *n_ptr;
2400         struct tipc_node *found_node = 0;
2401         int i;
2402
2403         *bearer_id = 0;
2404         rcu_read_lock();
2405         list_for_each_entry_rcu(n_ptr, &tipc_node_list, list) {
2406                 tipc_node_lock(n_ptr);
2407                 for (i = 0; i < MAX_BEARERS; i++) {
2408                         l_ptr = n_ptr->links[i];
2409                         if (l_ptr && !strcmp(l_ptr->name, link_name)) {
2410                                 *bearer_id = i;
2411                                 found_node = n_ptr;
2412                                 break;
2413                         }
2414                 }
2415                 tipc_node_unlock(n_ptr);
2416                 if (found_node)
2417                         break;
2418         }
2419         rcu_read_unlock();
2420
2421         return found_node;
2422 }
2423
2424 /**
2425  * link_value_is_valid -- validate proposed link tolerance/priority/window
2426  *
2427  * @cmd: value type (TIPC_CMD_SET_LINK_*)
2428  * @new_value: the new value
2429  *
2430  * Returns 1 if value is within range, 0 if not.
2431  */
2432 static int link_value_is_valid(u16 cmd, u32 new_value)
2433 {
2434         switch (cmd) {
2435         case TIPC_CMD_SET_LINK_TOL:
2436                 return (new_value >= TIPC_MIN_LINK_TOL) &&
2437                         (new_value <= TIPC_MAX_LINK_TOL);
2438         case TIPC_CMD_SET_LINK_PRI:
2439                 return (new_value <= TIPC_MAX_LINK_PRI);
2440         case TIPC_CMD_SET_LINK_WINDOW:
2441                 return (new_value >= TIPC_MIN_LINK_WIN) &&
2442                         (new_value <= TIPC_MAX_LINK_WIN);
2443         }
2444         return 0;
2445 }
2446
2447 /**
2448  * link_cmd_set_value - change priority/tolerance/window for link/bearer/media
2449  * @name: ptr to link, bearer, or media name
2450  * @new_value: new value of link, bearer, or media setting
2451  * @cmd: which link, bearer, or media attribute to set (TIPC_CMD_SET_LINK_*)
2452  *
2453  * Caller must hold RTNL lock to ensure link/bearer/media is not deleted.
2454  *
2455  * Returns 0 if value updated and negative value on error.
2456  */
2457 static int link_cmd_set_value(const char *name, u32 new_value, u16 cmd)
2458 {
2459         struct tipc_node *node;
2460         struct tipc_link *l_ptr;
2461         struct tipc_bearer *b_ptr;
2462         struct tipc_media *m_ptr;
2463         int bearer_id;
2464         int res = 0;
2465
2466         node = tipc_link_find_owner(name, &bearer_id);
2467         if (node) {
2468                 tipc_node_lock(node);
2469                 l_ptr = node->links[bearer_id];
2470
2471                 if (l_ptr) {
2472                         switch (cmd) {
2473                         case TIPC_CMD_SET_LINK_TOL:
2474                                 link_set_supervision_props(l_ptr, new_value);
2475                                 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2476                                                      new_value, 0, 0);
2477                                 break;
2478                         case TIPC_CMD_SET_LINK_PRI:
2479                                 l_ptr->priority = new_value;
2480                                 tipc_link_proto_xmit(l_ptr, STATE_MSG, 0, 0,
2481                                                      0, new_value, 0);
2482                                 break;
2483                         case TIPC_CMD_SET_LINK_WINDOW:
2484                                 tipc_link_set_queue_limits(l_ptr, new_value);
2485                                 break;
2486                         default:
2487                                 res = -EINVAL;
2488                                 break;
2489                         }
2490                 }
2491                 tipc_node_unlock(node);
2492                 return res;
2493         }
2494
2495         b_ptr = tipc_bearer_find(name);
2496         if (b_ptr) {
2497                 switch (cmd) {
2498                 case TIPC_CMD_SET_LINK_TOL:
2499                         b_ptr->tolerance = new_value;
2500                         break;
2501                 case TIPC_CMD_SET_LINK_PRI:
2502                         b_ptr->priority = new_value;
2503                         break;
2504                 case TIPC_CMD_SET_LINK_WINDOW:
2505                         b_ptr->window = new_value;
2506                         break;
2507                 default:
2508                         res = -EINVAL;
2509                         break;
2510                 }
2511                 return res;
2512         }
2513
2514         m_ptr = tipc_media_find(name);
2515         if (!m_ptr)
2516                 return -ENODEV;
2517         switch (cmd) {
2518         case TIPC_CMD_SET_LINK_TOL:
2519                 m_ptr->tolerance = new_value;
2520                 break;
2521         case TIPC_CMD_SET_LINK_PRI:
2522                 m_ptr->priority = new_value;
2523                 break;
2524         case TIPC_CMD_SET_LINK_WINDOW:
2525                 m_ptr->window = new_value;
2526                 break;
2527         default:
2528                 res = -EINVAL;
2529                 break;
2530         }
2531         return res;
2532 }
2533
2534 struct sk_buff *tipc_link_cmd_config(const void *req_tlv_area, int req_tlv_space,
2535                                      u16 cmd)
2536 {
2537         struct tipc_link_config *args;
2538         u32 new_value;
2539         int res;
2540
2541         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_CONFIG))
2542                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2543
2544         args = (struct tipc_link_config *)TLV_DATA(req_tlv_area);
2545         new_value = ntohl(args->value);
2546
2547         if (!link_value_is_valid(cmd, new_value))
2548                 return tipc_cfg_reply_error_string(
2549                         "cannot change, value invalid");
2550
2551         if (!strcmp(args->name, tipc_bclink_name)) {
2552                 if ((cmd == TIPC_CMD_SET_LINK_WINDOW) &&
2553                     (tipc_bclink_set_queue_limits(new_value) == 0))
2554                         return tipc_cfg_reply_none();
2555                 return tipc_cfg_reply_error_string(TIPC_CFG_NOT_SUPPORTED
2556                                                    " (cannot change setting on broadcast link)");
2557         }
2558
2559         res = link_cmd_set_value(args->name, new_value, cmd);
2560         if (res)
2561                 return tipc_cfg_reply_error_string("cannot change link setting");
2562
2563         return tipc_cfg_reply_none();
2564 }
2565
2566 /**
2567  * link_reset_statistics - reset link statistics
2568  * @l_ptr: pointer to link
2569  */
2570 static void link_reset_statistics(struct tipc_link *l_ptr)
2571 {
2572         memset(&l_ptr->stats, 0, sizeof(l_ptr->stats));
2573         l_ptr->stats.sent_info = l_ptr->next_out_no;
2574         l_ptr->stats.recv_info = l_ptr->next_in_no;
2575 }
2576
2577 struct sk_buff *tipc_link_cmd_reset_stats(const void *req_tlv_area, int req_tlv_space)
2578 {
2579         char *link_name;
2580         struct tipc_link *l_ptr;
2581         struct tipc_node *node;
2582         unsigned int bearer_id;
2583
2584         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2585                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2586
2587         link_name = (char *)TLV_DATA(req_tlv_area);
2588         if (!strcmp(link_name, tipc_bclink_name)) {
2589                 if (tipc_bclink_reset_stats())
2590                         return tipc_cfg_reply_error_string("link not found");
2591                 return tipc_cfg_reply_none();
2592         }
2593         node = tipc_link_find_owner(link_name, &bearer_id);
2594         if (!node)
2595                 return tipc_cfg_reply_error_string("link not found");
2596
2597         tipc_node_lock(node);
2598         l_ptr = node->links[bearer_id];
2599         if (!l_ptr) {
2600                 tipc_node_unlock(node);
2601                 return tipc_cfg_reply_error_string("link not found");
2602         }
2603         link_reset_statistics(l_ptr);
2604         tipc_node_unlock(node);
2605         return tipc_cfg_reply_none();
2606 }
2607
2608 /**
2609  * percent - convert count to a percentage of total (rounding up or down)
2610  */
2611 static u32 percent(u32 count, u32 total)
2612 {
2613         return (count * 100 + (total / 2)) / total;
2614 }
2615
2616 /**
2617  * tipc_link_stats - print link statistics
2618  * @name: link name
2619  * @buf: print buffer area
2620  * @buf_size: size of print buffer area
2621  *
2622  * Returns length of print buffer data string (or 0 if error)
2623  */
2624 static int tipc_link_stats(const char *name, char *buf, const u32 buf_size)
2625 {
2626         struct tipc_link *l;
2627         struct tipc_stats *s;
2628         struct tipc_node *node;
2629         char *status;
2630         u32 profile_total = 0;
2631         unsigned int bearer_id;
2632         int ret;
2633
2634         if (!strcmp(name, tipc_bclink_name))
2635                 return tipc_bclink_stats(buf, buf_size);
2636
2637         node = tipc_link_find_owner(name, &bearer_id);
2638         if (!node)
2639                 return 0;
2640
2641         tipc_node_lock(node);
2642
2643         l = node->links[bearer_id];
2644         if (!l) {
2645                 tipc_node_unlock(node);
2646                 return 0;
2647         }
2648
2649         s = &l->stats;
2650
2651         if (tipc_link_is_active(l))
2652                 status = "ACTIVE";
2653         else if (tipc_link_is_up(l))
2654                 status = "STANDBY";
2655         else
2656                 status = "DEFUNCT";
2657
2658         ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
2659                             "  %s  MTU:%u  Priority:%u  Tolerance:%u ms"
2660                             "  Window:%u packets\n",
2661                             l->name, status, l->max_pkt, l->priority,
2662                             l->tolerance, l->queue_limit[0]);
2663
2664         ret += tipc_snprintf(buf + ret, buf_size - ret,
2665                              "  RX packets:%u fragments:%u/%u bundles:%u/%u\n",
2666                              l->next_in_no - s->recv_info, s->recv_fragments,
2667                              s->recv_fragmented, s->recv_bundles,
2668                              s->recv_bundled);
2669
2670         ret += tipc_snprintf(buf + ret, buf_size - ret,
2671                              "  TX packets:%u fragments:%u/%u bundles:%u/%u\n",
2672                              l->next_out_no - s->sent_info, s->sent_fragments,
2673                              s->sent_fragmented, s->sent_bundles,
2674                              s->sent_bundled);
2675
2676         profile_total = s->msg_length_counts;
2677         if (!profile_total)
2678                 profile_total = 1;
2679
2680         ret += tipc_snprintf(buf + ret, buf_size - ret,
2681                              "  TX profile sample:%u packets  average:%u octets\n"
2682                              "  0-64:%u%% -256:%u%% -1024:%u%% -4096:%u%% "
2683                              "-16384:%u%% -32768:%u%% -66000:%u%%\n",
2684                              s->msg_length_counts,
2685                              s->msg_lengths_total / profile_total,
2686                              percent(s->msg_length_profile[0], profile_total),
2687                              percent(s->msg_length_profile[1], profile_total),
2688                              percent(s->msg_length_profile[2], profile_total),
2689                              percent(s->msg_length_profile[3], profile_total),
2690                              percent(s->msg_length_profile[4], profile_total),
2691                              percent(s->msg_length_profile[5], profile_total),
2692                              percent(s->msg_length_profile[6], profile_total));
2693
2694         ret += tipc_snprintf(buf + ret, buf_size - ret,
2695                              "  RX states:%u probes:%u naks:%u defs:%u"
2696                              " dups:%u\n", s->recv_states, s->recv_probes,
2697                              s->recv_nacks, s->deferred_recv, s->duplicates);
2698
2699         ret += tipc_snprintf(buf + ret, buf_size - ret,
2700                              "  TX states:%u probes:%u naks:%u acks:%u"
2701                              " dups:%u\n", s->sent_states, s->sent_probes,
2702                              s->sent_nacks, s->sent_acks, s->retransmitted);
2703
2704         ret += tipc_snprintf(buf + ret, buf_size - ret,
2705                              "  Congestion link:%u  Send queue"
2706                              " max:%u avg:%u\n", s->link_congs,
2707                              s->max_queue_sz, s->queue_sz_counts ?
2708                              (s->accu_queue_sz / s->queue_sz_counts) : 0);
2709
2710         tipc_node_unlock(node);
2711         return ret;
2712 }
2713
2714 struct sk_buff *tipc_link_cmd_show_stats(const void *req_tlv_area, int req_tlv_space)
2715 {
2716         struct sk_buff *buf;
2717         struct tlv_desc *rep_tlv;
2718         int str_len;
2719         int pb_len;
2720         char *pb;
2721
2722         if (!TLV_CHECK(req_tlv_area, req_tlv_space, TIPC_TLV_LINK_NAME))
2723                 return tipc_cfg_reply_error_string(TIPC_CFG_TLV_ERROR);
2724
2725         buf = tipc_cfg_reply_alloc(TLV_SPACE(ULTRA_STRING_MAX_LEN));
2726         if (!buf)
2727                 return NULL;
2728
2729         rep_tlv = (struct tlv_desc *)buf->data;
2730         pb = TLV_DATA(rep_tlv);
2731         pb_len = ULTRA_STRING_MAX_LEN;
2732         str_len = tipc_link_stats((char *)TLV_DATA(req_tlv_area),
2733                                   pb, pb_len);
2734         if (!str_len) {
2735                 kfree_skb(buf);
2736                 return tipc_cfg_reply_error_string("link not found");
2737         }
2738         str_len += 1;   /* for "\0" */
2739         skb_put(buf, TLV_SPACE(str_len));
2740         TLV_SET(rep_tlv, TIPC_TLV_ULTRA_STRING, NULL, str_len);
2741
2742         return buf;
2743 }
2744
2745 /**
2746  * tipc_link_get_max_pkt - get maximum packet size to use when sending to destination
2747  * @dest: network address of destination node
2748  * @selector: used to select from set of active links
2749  *
2750  * If no active link can be found, uses default maximum packet size.
2751  */
2752 u32 tipc_link_get_max_pkt(u32 dest, u32 selector)
2753 {
2754         struct tipc_node *n_ptr;
2755         struct tipc_link *l_ptr;
2756         u32 res = MAX_PKT_DEFAULT;
2757
2758         if (dest == tipc_own_addr)
2759                 return MAX_MSG_SIZE;
2760
2761         n_ptr = tipc_node_find(dest);
2762         if (n_ptr) {
2763                 tipc_node_lock(n_ptr);
2764                 l_ptr = n_ptr->active_links[selector & 1];
2765                 if (l_ptr)
2766                         res = l_ptr->max_pkt;
2767                 tipc_node_unlock(n_ptr);
2768         }
2769         return res;
2770 }
2771
2772 static void link_print(struct tipc_link *l_ptr, const char *str)
2773 {
2774         struct tipc_bearer *b_ptr;
2775
2776         rcu_read_lock();
2777         b_ptr = rcu_dereference_rtnl(bearer_list[l_ptr->bearer_id]);
2778         if (b_ptr)
2779                 pr_info("%s Link %x<%s>:", str, l_ptr->addr, b_ptr->name);
2780         rcu_read_unlock();
2781
2782         if (link_working_unknown(l_ptr))
2783                 pr_cont(":WU\n");
2784         else if (link_reset_reset(l_ptr))
2785                 pr_cont(":RR\n");
2786         else if (link_reset_unknown(l_ptr))
2787                 pr_cont(":RU\n");
2788         else if (link_working_working(l_ptr))
2789                 pr_cont(":WW\n");
2790         else
2791                 pr_cont("\n");
2792 }